Cryptopolitan

Papertrade is the latest onchain derivatives platform that’s making moves. It went live on HyperEVM, Hyperliquid’s smart contract layer, on October 10. Just four days later, its revenue has hit $37.03 million, according to DefiLlama. Pump.fun generated $16.06 million in seven day revenue while Hyperliquid made $14.56 million. Combine the two and Papertrade is already printing more than two of the biggest names in the space.
On the 24 hour timeframe, Papertrade ranks third among the over 2600 protocols DefiLlama tracks, behind only Tether and Circle.
Every Papertrade Trade Runs Against a Shared Pool
On most exchanges, one trader buys, another sells and the exchange charges a small fee for matching them. Papertrade works in a different way as there is no second trader to be matched. Users bet on whether BTC or ETH goes up or down and a shared pool of USDC takes the other side.
When a trader wins, the pool pays out. On the other hand, when a trader is on the losing side, the money goes into the pool and the trader receives newly minted PAPER tokens, starting at 100 PAPER per $1 lost. Papertrade’s official docs also state that the token has no team, VC or airdrop allocation, so losing is the only way the tokens come into existence. Once the pool tops $5 million, everything above that goes to people staking PAPER. As of now, the tokens can’t be sold or sent between wallets.
The platform offers leverage of up to 1000x and this is why losses can pile up very fast. With that amount of leverage, a $100 deposit becomes a $100,000 position and just a 0.1% move the wrong way can wipe out trades.
Most of Papertrade’s revenue is that overflow of trader losses. The $37 million figure is mostly a record of how much users have lost since Saturday.
One Wallet Moved the ETH Price and Took $1.28 Million
Papertrade does not run an order book and it takes the prices of BTC and ETH from Hyperliquid’s. It uses the midpoint between the best price buyers offer and the best price sellers ask. So in this sense, when hyperliquid’s price fluctuates, you move Papertrade’s as well.
Wallet 0x2025…840e needed less than a day to figure this out. Between 6:01 and 6:02 PM ET on October 10, the wallet bought around 6,060 ETH on Hyperliquid. At the same time, this wallet was already open with massive leveraged longs on Papertrade. ETH went up by around $3 and while a 0.12% barely registers for most traders, a position with a 1000x leverage makes this a sizable move. The wallet sold, then later flipped to shorts and did the reverse.
The bill on Hyperliquid was roughly $52,000 in fees on $231 million of ETH. The payout on Papertrade was $1.28 million. Papertrade has since cut its open position limits.
New Losses Pay Out To Earlier Losers
PAPER changes who the house is. Stakers collect the pool’s surplus, and every staker got their tokens by taking a losing trade. Once the pool tops $5 million, new losses flow back to traders who lost before.
That isn’t a refund. A trader who loses $1,000 gets PAPER that can’t be sold yet, though staking it does pay out in USDC. The mint rate also drops once the pool passes $2 million, so later losers get less PAPER per dollar than early ones.
The design needs a steady flow of new losses. When the pool can’t cover a winning trade, Papertrade queues the unpaid profit and settles it from future traders’ losses. If losses slow, winners wait longer and stakers earn less.
The smartest crypto minds already read our newsletter. Want in? Join them.

An OpenAI engineer has been accused of hiding a recording device in an executive conference room in San Francisco and now is facing charges from the federal authorities.
The surprising turn is that AI technology had supposedly assisted the individual in his preparation, while OpenAI said that its models helped investigate the case.
What prosecutors say Alfozan did
Prosecutors said that Abdulrahman Alfozan, aged 35, hid the recording device in toilet paper and attached it to the USB port of a television set on September 28, 2026. OpenAI’s security discovered it during a search on September 30, according to the Department of Justice.
The gadget is identified as a $13.99 Seeed XIAO ESP32-S3 Sense board that comes with an encrypted microSD storage, according to WIRED. Authorities reviewed the surveillance clips. They discovered a second recorder in an unattended bag containing Alfozan’s identification on October 5.
After being arrested on October 9, he was brought to court on October 13. If he is found guilty, he will spend five years in jail and pay the fine of $250,000. According to the New York Times, another hearing is scheduled for October 19. The prosecution did not provide any information regarding the motive.
The chatbot trail, and OpenAI’s AI-assisted response
Prosecutors claim that Alfozan used a workplace productivity agent to search executive meeting rooms and schedules. Also, a chatbot account linked to his spouse was used to check for around-the-clock recording, access to Wi-Fi, as well as ability for live listening and SD card encryption.
OpenAI claimed that its models allowed security staff to understand the device and identify who planted it. But AI was only part of the investigation. Camera footage, chat records and physical evidence also helped in the investigation.
The company has found no evidence to suggest that model weights, source code of its models, or customer data were accessed or stolen without permission.
What a single insider can cost
If confidential data was compromised, the situation would have significantly worsened and become more expensive. IBM states that data breaches cost businesses an average of $4.99 million globally. Organizations that use security AI and automation tools have reported breach costs that are $1.93 million lower than companies that do not.
Insider attacks create more financial risks. The Ponemon Institute and DTEX estimate that the average cost of containing an insider breach incident is $247,587. For companies that took over 90 days to contain an incident, the average cost is $21.9 million. The figures in this report reflect the average costs in the industry, not the losses attributed to OpenAI.
Why budgets and AI oversight are climbing
PwC found 84% of surveyed security and finance leaders expect cyber budgets to rise, while 58% of security leaders prioritize AI spending.
Gartner projects AI governance-platform spending will rise from $492 million in 2026 to more than $1 billion by 2030. BCG found nearly half of companies already see value from AI, yet only 5% have complete controls for autonomous agents.
OpenAI Insider Threat Case: Cybersecurity Costs and AI Security Spending
A separate fight over watching the models
Cryptopolitan reported that three former OpenAI researchers have separately questioned how the company monitors advanced models and called for independent safety reviews. That debate concerns model behavior, not the alleged bugging. Both show how quickly trust and oversight are becoming business issues for AI companies.
If you're reading this, you’re already ahead. Stay there with our newsletter.

The U.S. Department of the Treasury imposed more than $1.21 billion in penalties on Bank of America, the country’s second-largest bank, for breaches of anti-money-laundering and sanctions regulations.
According to the Financial Crimes Enforcement Network (FinCEN), the Bank faced penalties for failing to adequately address or report major national security risks, including money-laundering networks associated with China and illicit financial transactions along the U.S. southwest border. To date, these fines are the third-largest amount ever issued by the department.
BoA will pay $750 million to the U.S. Office of the Comptroller of the Currency
The $1.2 billion penalty includes a historic $750 million payout to the U.S. Office of the Comptroller of the Currency (OCC), marking the agency’s largest fine to date. Additionally, the Office of Foreign Assets Control (OFAC) secured an $11.6 million settlement over the bank’s apparent violations of sanctions targeting Iran, Russia, narcotics, and international terrorism.
In its press release, the OCC said it found that the institution’s severely flawed tracking system allowed trillions of dollars to slip through without proper oversight. A flaw that forced the bank to file late red-flag reports on billions of dollars, much of it tied to suspected money laundering.
Furthermore, it claimed the bank was aware for years that potentially illicit transactions were not being escalated for investigation, but failed to remedy this critical compliance failure. Additionally, it noted that the bank’s compliance leadership and staff were underqualified to manage a robust anti-money laundering program.
A lack of targeted training on how to properly investigate and report suspicious transactions ultimately caused the bank to submit flawed and overdue compliance reports to regulators, the agency said.
Suspicious activity reports are an important part of the U.S. financial system’s efforts to combat illicit finance. Banks submit these reports to FinCEN when transactions raise concerns about possible illegal activity.
Regulators and law enforcement agencies can use the information to identify suspicious financial networks, investigate potential crimes, and trace the movement of funds. Delayed or incomplete reports can make it harder for authorities to detect patterns across multiple transactions or institutions, particularly when suspected activity involves cross-border transfers and complex money-laundering networks.
As noted by the Treasury, Bank of America has admitted to violating the 1970 Bank Secrecy Act, which requires banks to help authorities prevent money laundering. The bank contended, “Financial crime compliance is serious, and we are responsible for getting it right. We worked with regulators to reach these penalty resolutions stemming from violations cited in the previously reported 2024 (Comptroller of the Currency) Consent Order and FinCEN’s investigation.”
The Bank Secrecy Act requires financial institutions to maintain effective anti-money laundering programs, keep appropriate transaction records, and report suspicious financial activity to authorities. Such measures help law enforcement identify money laundering, terrorist financing, and other financial crimes.
Financial institutions that do not detect or report suspicious transactions can leave gaps that criminals exploit to move illicit funds through the financial system. The Bank of America case demonstrates the importance of effective monitoring systems and timely reporting in helping regulators identify financial risks.
It added that so far, they’ve spent billions of dollars on technology and hired thousands of people to stop financial crime. Nonetheless, it acknowledged there’s still more to be done and is working to resolve the remaining issues flagged by regulators.
Also, regarding the fines, US Treasury Secretary Scott Bessent noted that the department’s action shows that even the country’s biggest banks are not exempt from legal accountability.
Binance faced a $4 billion penalty
Only two fines in Treasury history top this latest one: the $4 billion Binance penalty in 2023 and TD Bank’s $3 billion settlement in 2024.
As reported by Cryptopolitan, in 2023, Changpeng Zhao stepped down from his role as chief executive as Binance acknowledged its failures. The crypto exchange admitted it violated the Bank Secrecy Act and ignored sanctions programs by failing to implement proper systems to track suspicious transactions.
At the time, the then U.S. Attorney General Merrick Garland even sharply criticized the exchange, emphasizing that employing novel technology to bypass legal frameworks does not make a firm a market disruptor, but rather a criminal.
Binance was subjected to a $4 billion penalty and a five-year monitoring period. Later, Binance founder Changpeng Zhao was sentenced to four months in prison for allowing his exchange to become a haven for terror financing, drug trafficking, and child abuse material.
In a similar vein, Canada-based TD Bank agreed to a $3 billion penalty after structural deficiencies enabled systemic money laundering over a multi-year period. “TD Bank created an environment that allowed financial crime to flourish. By making its services convenient for criminals, it became one,” Garland had noted.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Binance will shut down its Bitcoin mining pool on November 10, 2026, at 07:00 UTC. The exchange plans to expand its AI tools and traditional assets trading. Miners will have a few weeks to move to other mining pools.
What closes on November 10
Binance said in a notice released on October 13 that Pool, Cloud Mining, Pool Savings and Transaction Accelerator will shut down, together with Stratum connections, APIs, and email and Telegram alerts.
The exchange warned in its official announcement:
Hashrate submissions and mining shares submitted after this time will no longer be accepted, processed, or credited.
Before the closure, miners must transfer to another pool and close their Pool Savings accounts. Pending rewards or refunds will be processed based on existing product terms and after checking for compliance.
Users will still be able to access their transaction history until November 24, 07:00 UTC. Binance said the closure was part of a broader product review. However, it did not explain the financial reasons behind its decision.
From ninth place to the bottom of the dashboard
Binance Pool started its operations in April 2020, advocating the Full Pay-Per-Share model for Bitcoin miners.
Hashrate Index rank for December 31, 2025 had it at number 9 with a hashrate of 26 EH/s and a market share of 2.3%. Its dashboard for October 14, 2026, ranked Binance Pool on the 11th position with a hashrate of about 9.1 EH/s and a market share of 0.85%, in comparison with total network hashrate of approximately 1,013 EH/s.
The numbers involve different times and measurement durations and thus cannot be interpreted as the performance history.
Where the hashrate goes next
It is possible for mining rigs to link up with an alternative pool while remaining a part of Bitcoin’s network. However, concentration might become a problem. Foundry USA make up about 25% of recently mined blocks, while AntPool hold around 22%. In case Binance miners became part of those pools, those shares would increase more.
According to calculations made by Luxor, hashprice in September was $39.33 per PH/s per day. This figure refers to mining economics in general, as Binance Pool’s profitability remains unknown.
Trading tools over mining rigs
Meanwhile, Binance released its free AI market assistant on October 12 after a beta involving roughly 150,000 users. It provides market guidance, not autonomous trading.
Binance has also expanded into U.S. equities through Nest Trading and announced plans for physically settled options on more than 1,000 U.S. stocks and ETFs. Its reported TradFi perpetual-futures volume climbed from $29.5 billion in January to $433.4 billion in August, according to The Block.
It was also reported by Cryptopolitan that Binance commanded 25.3% liquidity in Bitcoin on the exchanges studied. On another note, CoinGecko estimated that there was $1.45 trillion in traditional-asset trading across six crypto exchanges during the first half of 2026.
Binance Mining Pool Rank Drops to 11th as TradFi Trading Volume Reaches $433.4 Billion
Bitcoin and AI, close but not the same trade
Perception found that one in eight Bitcoin mentions in the third quarter also referenced AI, up from one in 37 in 2022. Much of this growing interest comes from Bitcoin miners turning to AI data centers. They already have access to the power and infrastructure these facilities need.
So far, however, the shift has not paid off for everyone. Bitcoin gained 43% during the quarter, while five miners with major AI deals saw their stocks fall 30%. Their share prices now track Nvidia more closely than Bitcoin.
Binance is taking a different path. While some miners are betting on AI infrastructure, the exchange is closing its mining pool and expanding its trading services. Both developments show how the crypto industry is exploring new sources of growth beyond Bitcoin mining.
The smartest crypto minds already read our newsletter. Want in? Join them.

An organized campaign against scam operations in West Africa has resulted in the arrest of 28 people in Ghana and Nigeria. The Justice Department said that there is about $11 million in cryptocurrencies that can possibly be seized.
The more significant move, however, was the efforts toward shutting down the recipients’ infrastructure and financial operations.
Going after the accounts, not just the operators
The campaign called “Joint Disruption Week” started on October 6 at the head office of the UK’s National Crime Agency in London. The investigators joined forces with such companies as Meta, Google, Microsoft, OpenAI, Binance, Coinbase, and many others. Together they managed to take down over 3.7 million accounts, Pages and Groups, in addition to sharing useful clues on over 30 suspected scam centers.
“We are not simply trying to stop individual scams,” FBI official Matthew Floyd told Axios. “We intend to dismantle the criminal enterprises behind them.”
US Attorney Jeanine Ferris Pirro stated that a piece of information from Meta made it possible to organize this campaign.
The numbers behind the raids
Ghanaian authorities arrested 19 people and seized 303 devices; Nigerian officers arrested nine and seized 53. In Ghana, officials also found 117 trafficking victims forced to conduct scams. They await repatriation to Nigeria.
Importantly, it must be noted that it was pointed out that $11 million may potentially be frozen, but has not been confirmed to be seized. Coinbase stated that they have found more than 65 active victims and prevented more than $415,000 from being lost.
According to FBI Director Kash Patel, agents “arrested dozens of criminal actors, seized hundreds of devices, identified at least $11 million in potentially freezable funds, and truly saved lives.”
A smaller, messier version of the Southeast Asia problem
Networks in West Africa are less concentrated and more diffused than in Southeast Asia. Their schemes run from fake jobs and romance scams to police impersonation and bogus recovery offers. This makes enforcement harder since investigators must target multiple accounts and payment channels rather than a handful of centralized operations.
There is a significant financial risk involved. Chainalysis estimates that crypto scams could have stolen more than $17 billion, while impersonation fraud increased by over 1,400% in 2025. While the $11 million identified in West Africa is significant, it is much smaller than the amount reported in previous American operations targeted at networks in Southeast Asia.
Context is necessary regarding the comparison. The DOJ has cracked down on a series of Chinese-led fraud schemes in Burma and Cambodia last April, targeting everything from bogus investment promises to trafficking rings. By that point, authorities had seized more than $700 million in total.
In September, investigators targeted Xinbi, a Chinese-language site providing money laundering services to scammers. This one operation grabbed an additional $52 million, taking total funds seized to around $938 million. But unlike those funds, the $11 million from West Africa is only potentially freezable.
$11M Flagged in West Africa vs. $938M Restrained in Southeast Asia-Linked Scam Cases
The hard part is getting the money back
A September FinCEN alert explains how scammers move proceeds through layers of wallets and intermediaries. TRM Labs estimates Chinese-language laundering networks processed over $100 billion in 2025.
Speed presents another barrier. Earlier, Cryptopolitan covered a study of 2,955 Tether freezes that indicated an average gap of more than two hours between the initiation and implementation of the freeze, allowing the funds to escape. FATF findings showed some blind spots offshore exchanges may face, including a Nigerian case of frozen wallet containing around $600 million in funds.
What it means for exchanges and the wider market
Binance, Coinbase and stablecoin issuers face growing pressure to flag suspicious flows, file reports and respond quickly, raising compliance costs.
Faster intervention could strengthen confidence in crypto markets. But flagged funds are not recovered funds, and tougher controls must avoid disrupting legitimate users.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Google has officially released Gemini 4 Argon, its newest frontier AI model, to developers and enterprises worldwide, pricing the model below OpenAI and Anthropic as it chases more business and developer customers.
Pricing aimed at undercutting rivals
Google’s launch blog lists introductory pricing at $2 per million input tokens and $10 per million output tokens, with cached inputs costing 95% less, or $0.10 per million tokens. The rates are set to remain in effect till November 30, after which they double.
This entry point sits below both main competitors, as Anthropic’s Claude Opus 5.5 charges $4 and $20 per million input and output tokens respectively, while OpenAI’s GPT-6 Astra charges $10 and $50. On a per-task basis, Mashable puts Argon at about $1.99 against the Artificial Analysis Intelligence Index, which is almost 60% of what GPT-6 Astra costs.
Developers and enterprises will assess Argon through the Gemini API in AI Studio and the Gemini platform. Google AI Pro and Ultra subscribers can use it right in the Gemini app and AI Mode in Search, while the Antigravity coding tool stays limited to Ultra accounts, according to Google’s post on X.
Gemini 4 Argon is available now for Google AI Ultra and Pro subscribers in @GeminiApp and AI Mode in Search, Ultra subscribers in @Antigravity, and to developers and enterprises via the Gemini API in @GoogleAIStudio and the Gemini platform.
Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.
— Google (@Google) October 13, 2026
Where Google’s Gemini Argon lands on benchmarks
Google says Argon tops the Vals Index, which measures performance across finance, coding, legal and tax tasks. However, the company’s own comparison table shows Argon leading in 14 of 18 results, tying with GPT-6 on cybersecurity and trailing in five.
On Artificial Analysis’ Intelligence Index, which tests models on standardized hardware, Argon scores 53 on its “High” setting. That ties with GPT-6 Astra and just about edges out GPT-6.1 Sol and Meta’s new Muse Wonder Max by a point.
It trails Claude Opus 5.5 and Claude Sonnet 5.5 by about five points though, while still marking a jump of 23 points over Gemini 3.1 Pro Preview, Google’s previous top model.
Early customers and a cyber carve-out
Google named Deloitte, LexisNexis, Ambience Healthcare, and Isomorphic Labs among the first firms testing Argon. The tech giant is also expanding its Fairwind Program to include hundreds more partners, giving vetted cybersecurity defenders access to a version of the model with its cyber safeguards removed.
Google claims its own engineering teams running Argon in Antigravity increased agentic code submissions by over 35% and reduced code rollbacks by almost a third.
The company also credited the model with helping Chrome engineers complete an architectural overhaul spanning hundreds of files in Chromium’s 40-million-line codebase.
Launch comes after rocky period at DeepMind
Google stated in September that it would no longer ship Gemini 3.5 Pro, a model CEO Sundar Pichai had once slated for June. Google later reorganized DeepMind after co-founder Demis Hassabis stepped aside and several Gemini leaders left the company.
Argon first surfaced on September 30 in a post by DeepMind SVP Koray Kavukcuoglu, who said access was initially limited to vetted cyber defenders in the Fairwind Program.
Google says Argon rejects requests involving cyberattacks or chemical, biological, radiological and nuclear threats. On Gray Swan’s prompt-injection test, it recorded a 0.7% attack success rate, compared with 1.0% for Opus 5.5 and 8.5% for GPT-6 Astra, with lower scores indicating stronger resistance.
It is worth keeping an eye out for November 30, as Argon’s introductory API pricing will then expire with its cost advantage starting to narrow.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Philip Lane, who serves as an Executive Board member of the European Central Bank (ECB), issued a warning on October 13 about the potential impact of artificial intelligence on the bargaining power of workers, even in times when workers do not experience mass job losses. Wage growth and labour income share will deteriorate before the employment numbers indicate that a shift has happened.
Lane sees weaker pay power before weaker payrolls
According to Lane, the option of substituting workers could actually work towards increasing the employers’ advantage. However, ECB does not see major changes in employment trends or precise productivity improvements across the economy.
The use of AI among employees in the euro zone rose from 26% in 2024 to 52% in 2026, and the usual AI user can save up to three hours a week, according to ECB’s research. However, savings made in time do not necessarily imply increases in productivity.
Who captures the productivity gains
Who gains from increased productivity? A survey conducted by PwC involving nearly 50,000 participants has revealed that nearly twice as many respondents indicated that their bargaining position was less secure than improved. The company’s “engine room”—56% of the interviewed professionals—has limited access to both AI and career advancement opportunities. As a result, the level of trust in management has dropped by 7%.
Still, jobs demanding AI skills are associated with higher salaries. According to PwC’s jobs study, AI-skilled jobs receive a 62% average premium, although logical causation is not established. As some jobs were “professionalised,” their growth in terms of numbers is twice as fast compared to the growth of “democratised” jobs, and their salaries increased since 2021 by 42%.
Evidence from the study in the European Economic Review proves that increased use of AI decreases the share of incomes going to labor.
Little sign of mass job losses, so far
There is limited evidence regarding mass layoffs. According to a Stanford study, unemployment in the US increased by 0.77 percentage points among workers most exposed to AI since 2022 compared to 0.85 percentage points for those least exposed to AI. This indicates a broader weakness of the labor market rather than a distinct shock produced by artificial intelligence.
Recent graduates face more challenges. Unemployment among them reached 5.6% in early 2026, a 1.6-point perspective increase over three years. According to Stanford, forecasts of AI accelerating mass layoffs, like the prediction made by Anthropic CEO Dario Amodei that AI would end up displacing half of all white-collar working positions and driving unemployment rate to 20% should be taken with caution.
AI Adoption Doubles as ECB Warns of Wage Risks: Jobs and Layoffs Compared
Europe’s funding gap and the AI valuation risk
Financing hurdles are also present in Europe. According to Lane, investment in the US witnessed growth that was almost double that of euro zone investments in the last two years. An ECB analysis also indicates a trend in which companies that rely on AI for their operations are moving into market financing. The higher cost of borrowing, driven by bank interest rates, creates further pressure.
At the same time, the October report from the IMF indicates that disappointing benefits from AI investments may lead to a market correction, exacerbated by financial leverage and concentrated bets. In the words of the IMF’s Athanasios Vamvakidis, “resilience is not safety.”
Why the Kroger comparison is not proof
Cryptopolitan reported that Kroger plans to cut around 3,000 of its 8,200 corporate jobs—about 37%—over three years. CEO Greg Foran announced the cuts October 12, weeks after the retailer unveiled its AI strategy. He cited streamlining and investment in store staff, not AI.
The timing raises questions. It does not prove AI caused the cuts—or that a wider layoff wave has begun.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Two class action lawsuits have been submitted against Elon Musk’s SpaceXAI in Chicago on accusations claiming that Grok scans people’s faces in any photos uploaded on the platform without their permission. The court hearings will help determine if facial scans produced by AI imaging technology qualify as biometric data.
The complaint says Grok maps faces without consent
The lawsuits, filed on October 7 and 12, allege that Grok scans facial features when users upload photos for analysis or editing. Cesar Padilla, a Cook County resident, claims the system records details such as the shape and positioning of the eyes, nose, lips and jaw. He also alleges that these scans can be traced to users’ accounts, with no publicly disclosed policy explaining when the data is deleted.
Why BIPA treats a scan differently from a photo
Padilla’s lawsuits differentiate a normal picture from the facial data derived from it. The Illinois Biometric Information Privacy Act (BIPA) requires companies to give written notice, secure signed consent, and disclose how long biometric data will be stored and when it will be disposed of. Violations are punishable with damages amounting to $1,000 for negligence or $5,000 for intentional or reckless conduct. However, the state law sets a maximum limit for damages for repeated scans.
Facebook settled for $650 million; BNSF for $75 million. Google and Otter.ai face similar claims. Paul Hastings attorneys Aaron Charfoos and Michelle Reed warned in an April advisory that “a new wave of biometrics litigation has arrived.” They added: “When they are used together, these exposures compound.”
xAI is already fighting Minnesota over Grok Imagine
As per the report by Cryptopolitan, Judge Donovan Frank refused to suspend Minnesota’s law banning AI nudification tools. The law stipulates fines of as much as $500,000 for every image, while xAI still pursues an appeal. Unlike the cases in Illinois, the law in Minnesota prohibits harmful AI-generated images, instead of restricting the collection of facial data. The punishments are awarded even to the service providers who had no idea that the pictures were created without consent.
The case quotes a report estimating Grok’s production of 3 million sexual images within a mere 11 days. Approximately 23,000 of the images were thought to show children. xAI claimed to have closed 52,222 accounts and reported to the National Center for Missing and Exploited Children (NCMEC) 73,604 cases.
The EU wrote different rules for the same technology
Since August 2, 2026, obligations provided in the AI Act are being enforced by the European Commission, including chatbot disclosures, deepfake labeling, and marking content in a machine-readable format. More than 180 organizations have joined in signing this code of transparency. Developers are now bound by new rules that regulate the lawful processing of images and the attribution of altered images to its authors.
Governance spending is climbing as rules fragment
Gartner expects spending on AI governance platforms to reach $492 million in 2026 and exceed $1 billion by 2030. AI regulations could affect 75% of the world’s economies, while effective governance tools could cut regulatory costs by 20%.
Meanwhile, BCG found that nearly half of companies are seeing returns from AI. Spending has doubled to 3.3% of revenue, yet only 5% have all the necessary safeguards for autonomous AI agents. Companies are investing more in AI, but oversight has yet to catch up.
Grok Lawsuits Highlight AI Privacy Costs, BIPA Settlements and Governance Spending
Researchers are building ways to keep faces off the server
A Purdue-led study introduces PrivateEdit, which masks identity-sensitive facial regions on-device before cloud editing and restores them locally afterward. Tests involving Gemini, Grok and LLaMA found facial-attribute inference often fell more than 50%.
For AI companies, privacy-by-design could reduce risk before courts settle the rules.
The smartest crypto minds already read our newsletter. Want in? Join them.

Evan Solomon, the Canadian Minister of Artificial Intelligence and Digital Innovation, announced on Tuesday that an additional $500 million over five years, on top of the previous $200 million, is now available through the Regional Artificial Intelligence Initiative, per the federal government.
The money brings the total federal funding for the program to $700 million. It is delivered through Canada’s regional development agencies, which connect companies in each part of the country to the program rather than routing applications through Ottawa.
The initiative aims to serve two categories of entities: companies seeking to incorporate artificial intelligence within their processes and Canadian artificial intelligence firms seeking clients and investment as well as scaling opportunities. The targeted industries under the initiative include health and life sciences, energy and natural resources, transportation, agriculture, and manufacturing and robotics.
Projects attracted $1.77 for every federal dollar approved
This is a 250% expansion to a pre-existing fund that started way back in 2024, and the government has gone ahead to show how the first $200 million was used up.
More than 218 projects have utilized AI in Canada through the program, either through adoption or commercialization, resulting in the creation or retention of over 1,100 jobs. These projects were able to attract an average of $1.77 from other sources for each dollar invested by Ottawa.
The multiplier is the important one, because the usual justification behind adoption programs is that public funds will release far more private investments. In this case, the multiplier works out to be only about 1.77, meaning that federal contribution is doing roughly a third of the lifting.
The next funding cycle has already started. For the province of Southern Ontario, the deadline to submit proposals is January 19, 2027. Projects that incur eligible expenditures prior to March 31, 2027 have a deadline of November 24, 2026.
Canada ties AI adoption funding to economic sovereignty
The language used by the minister extended beyond productivity to that of control.
The issue here is not whether the AI will transform our economic landscape. The real question is whether Canada will seize the chance. The objective is to ensure that the technology, talent, and economic benefit that we create here stays here and under Canadian jurisdiction.
– Evan Solomon
This sovereignty theme espoused by the minister permeates AI for All, the country’s strategy that was unveiled by Carney in Toronto on June 4, which sets out goals for $200 billion worth of economic growth and 250,000 job creation. There has also been some criticism. As mentioned earlier by Cryptopolitan, an official from the US Embassy stated in July that the United States backs sovereign AI in Canada provided that it doesn’t hinder superior American technology.
Moreover, he stated that the Canadian approach to imposing fines on AI companies for their non-compliance is contrary to the American view on corporate self-regulation.
Canada is trying to close its AI adoption gap
Ottawa committed $2 billion to compute through the Canadian Sovereign AI Compute Strategy in December 2024, including $705 million for a Canada-owned supercomputing system and $95 million for a secure computing facility. Budget 2025 added roughly $925 million over five years for sovereign public AI infrastructure. The target set by AI for All is to increase business AI usage from 12% to 60% by 2034, and this strategy recognizes that the gap lies among SMEs.
One of the key instruments that have been introduced in order to bridge this gap is the Regional Artificial Intelligence Initiative, which is the reason why it is provided by regional development organizations rather than being a single nationwide effort. The question whether $700 million is sufficient in order to make business adoption increase from 12% to 60% remains debatable.
This instrument also ensures broader access to Canada’s artificial intelligence economy, including Indigenous Peoples, youth, and underrepresented minorities. It complements other initiatives such as LIFT and the Compute Access Fund, both focusing on skill development and access to the computing resources required to create AI solutions.
The smartest crypto minds already read our newsletter. Want in? Join them.

On October 13, 2026, six House Democrats urged the Federal Aviation Administration (FAA) to pause the use of its new AI-assisted SMART tool. They want to receive answers regarding safety testing, controller training, and accountability before the technology is implemented across the US. Although over two dozen lawmakers have signed the letter, it does not prevent this technology from being adopted.
Who signed the letter and what they want
Representatives Alexandria Ocasio-Cortez, April McClain Delaney, Don Beyer, Kweisi Mfume, Suhas Subramanyam and Sarah Elfreth led the joint letter to Transportation Secretary Sean Duffy and FAA Administrator Bryan Bedford. Ocasio-Cortez’s office announced it on October 13.
SMART, which is an acronym for Strategic Management of Airspace, Routes and Trajectories, was initiated on September 21, 2026, and has begun operations in the airspace of Reagan National, Dulles, and Baltimore/Washington International airports. With plans for a nationwide rollout within three months, members of the Congress would like to understand how SMART was tested and if it is capable of dealing with adverse weather conditions, as well as whether controllers were consulted and trained on the system.
Ocasio-Cortez said, “Air traffic control leaves zero margin for error.”
McClain Delaney cited the collision in January 2025 near Reagan National, which resulted in the death of 67 people. She commented that “the FAA owes this region answers.”
What SMART actually does
The FAA’s September 21 announcement said SMART combines 200 data streams, including weather, flight paths and staffing, to predict congestion and help manage traffic.
The agency stressed that people remain in control:
“All of SMART’s recommendations are reviewed by FAA staff.”
However, the National Air Traffic Controllers Association has informed Nextgov/FCW that it “has not been involved in the design, testing or implementation of SMART.”
Later, Beyer mentioned that a discussion with Bedford helped calm some of his worries. Bedford stated that experienced air traffic controllers made SMART and that the system is operating at a Warrenton, Virginia-based radar approach facility, and not at the airport towers.
The modernization backdrop
The concerns come amid a much larger FAA overhaul. A September 14, 2026, GAO report found the modernization program lacked a reliable lifecycle cost estimate and an integrated master schedule.
Congress appropriated $12.5 billion in July 2025, mostly for phase one. The FAA estimates another $10.2 billion for phase two, excluding facilities.
The GAO’s findings concern the broader program, not SMART’s specific safety testing.
FAA SMART AI rollout timeline: 90-day expansion plan, $12.5B funding and $10.2B modernization estimate
Europe’s slower, documented path
Meanwhile, MarketsandMarkets expects aviation AI to grow from $1.76 billion in 2025 to $4.86 billion by 2030, representing 22.6% annual growth. Cryptopolitan previously reported a $2.67 trillion global AI spending forecast for 2026.
Europe is taking a more documented approach. Its aviation regulator released proposed AI guidance on June 3, 2026, covering safety assurance and human oversight. The guidance is not binding.
Similar US requirements could mean higher costs and slower deployments for AI vendors. For now, however, SMART’s expansion remains planned, and lawmakers are still seeking answers.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Trezor will stop its partnerships with some resellers starting January 1, 2027 due to the findings of its rival, Ledger, about altered wallets sold via CryptoBilis.
In a press release dated October 13, 2026, Trezor stated that “there is not a single piece of evidence that Trezor devices have been compromised.” The issue at hand is what happens to wallets between the time they leave Trezor’s premises and the time they reach customers.
What Sanders told the resellers
According to Chief Commercial Officer, Danny Sanders, the affected partners will receive neither new supply nor authorizations from January 1, 2027. The remaining partners are now required to follow stricter guidelines for sourcing, handling, and storing their supplies. Trezor has not revealed the number of resellers affected by the decision.
In the letter to resellers, Sanders has said:
“I am sorry. This decision says nothing about your effort or the way you represented us.”
The decision was made to improve oversight and does not mean that all cut-off resellers have done something wrong.
The implant Ledger found inside CryptoBilis wallets
According to Ledger Donjon on October 13, 2026, an unauthorized chip was discovered in wallets supplied by CryptoBilis. The chip recorded the 24-word recovery phrase that was shown during the initialization of the wallets.
The investigators elaborated:
“its microcontroller decodes the words and transmits them over LTE using a provisioned eSIM.”
The Secure Element, operating system, and transaction signing went unaffected by the implant. That means the wallet will continue to look normal while revealing the recovery phrase necessary for accessing the funds.
Loss estimates do vary. While Specter reported losses of more than $86 million, other reports have put the figure at around $93.2 million, while Yfarmx has given an estimate of $103 million. Ledger has not confirmed these numbers. Tether has frozen assets worth $10 million in USDT that are related to this case.
Cryptopolitan has earlier covered how the implant functions.
Why rivals are cutting the same reseller
OneKey has also canceled the authorization held by CryptoBilis. Users who obtained its products from March 1 to October 12, 2026, can apply for free replacements, including for activated wallets. Other brands sold by CryptoBilis have not confirmed thefts associated with them.
CryptoBilis customers who own unused Ledger wallets should avoid setting up the device. Those who are already using them should transfer their funds to a wallet where the recovery phrase has been newly generated from a device purchased directly from Ledger. They should contact official support and avoid disclosing recovery phrases.
Trezor has given detailed instructions to its potential wallet buyers. Check its list of authorized resellers to confirm the seller is legitimate. Upon delivery, check the tamper seals and conduct authenticity checks using Trezor Suite. If anything looks out of place, especially the firmware or setup instructions, contact Trezor. These precautions may help prevent possible hacks. But they cannot catch every hidden hardware modification.
What tighter distribution does to the wider market
Reports from Mordor Intelligence indicate the hardware-wallet market will evolve from $720 million in 2026 to $2.25 billion in 2031. Chainalysis ranks Brazil at the forefront of grassroots cryptocurrency adoption, pointing to growing demand for self-custody.
Hardware Wallet Market Growth 2026–2031: CryptoBilis Loss Estimates and Crypto Hack Statistics
Having fewer resellers could mean better security control, but it may also complicate the acquisition of wallets.
The stakes are extremely high already. TRM Labs has reported a total of 207 hacking cases involving cryptocurrencies and around $972 million stolen in the first half of 2026. CryptoBilis brought another threat to light: a hacker can hack a wallet even before the owner has had a chance to open the unit.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Around 400 scientists, including AI pioneers Geoffrey Hinton and Yoshua Bengio, have signed an open letter on Tuesday urging the creation of an international body with the power to enforce containment rules and require incident reporting for advanced AI systems capable of cyberattacks.
Supported by researchers from 34 countries, the letter calls for frontier AI models to be isolated from external networks from the outset, a practice known as air-gapping, and placed under independent oversight.
Reuters reported that the proposed body would set industry-wide containment standards and maintain a record of major security incidents, and signatories to information that would eventually be made public.
The scientists are urging organizations developing or testing cyber-capable AI models to adopt stringent containment measures immediately, without waiting for governments to establish formal oversight rules or pass legislation.
Voluntary pledges fall short of requirements
The researchers say voluntary safety pledges from AI developers do little to ensure accountability. Their appeal came about a month after Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed a White House agreement to adopt safety measures, with commitments that are not legally binding.
The scientists pointed to recent cases where AI models escaped poorly secured test environments and accessed external systems, claiming these incidents revealed the shortcomings in the efforts to keep AI technology under control.
Hinton and Bengio won the 2018 Turing Award alongside Yann LeCun, Meta’s chief AI scientist. LeCun declined to add his name, maintaining that the incidents cited in the letter stem from engineering shortcomings and not any looming existential threat from AI, according to Reuters.
Scientists warn there is less than a year to act
The call from the scientists adds to mounting warnings from within the AI industry. After a closed-door Capitol Hill briefing in September, Hinton told reporters that lawmakers might have just a year to put safeguards in place before AI moves well beyond human control.
Hinton and other senior researchers from OpenAI, Anthropic and Microsoft warned in a paper published earlier in October that AI systems capable of automating their own development could trigger an “intelligence explosion,” condensing years of progress into months or less. They urged governments to put oversight in place before such levels of rapid AI development happen.
Regulators still seem unsure and divided, though, on how much oversight AI really needs right now. EU tech chief Henna Virkkunen told Reuters on October 9 that the bloc’s AI Act already provides adequate safeguards against rogue AI systems. Two days later, the Trump administration also lifted restrictions on U.S. AI labs providing models to Britain’s AI Security Institute for any form of early safety testing.
The smartest crypto minds already read our newsletter. Want in? Join them.

Cipher Digital (NASDAQ: CIFR) has announced it intends to sell $850 million of zero-coupon convertible notes that are set to mature in 2032. The announcement was made on Tuesday, October 13, barely 3 weeks after the company agreed to absorb cost overruns to the tune of $359.3 million at its Barber Lake data center in Texas.
Cipher’s decision will serve as an example of how former Bitcoin miners now pay for their AI buildouts. When the cost of construction goes up, they segue to the convertible market.
Who pays for the Barber Lake overruns?
The cost overruns that Cipher is currently paying can be traced back to September 25. Cipher announced on the 25th of September that the contracted life of Barber Lake would be doubled to 20 years.
This caused contracted revenue to spike from $3.8 billion to over $9 billion. The extension of the Barber Lake Lease is coupled with a binding agreement with a leading AI lab. The 10-year commitment is meant to generate an additional $5.2 billion in incremental contracted revenue.
The agreements also featured changes in the party responsible for cost overruns. The company published a reimbursement framework in September 2026. Cipher will absorb the initial $359.3 million of cost overrun, while the tenant handles 50% above the original budget, and only above that line does the tenant cover half, repaid as extra rent over the full 20-year term.
No coupon, a 2030 escape hatch
The notes will mature on the 1st of November, 2032 and will not carry regular interest. Cipher plans to sell them on the private market under Rule 144A, but only to institutional buyers that qualify.
The first set of buyers will have a 13-day option to buy up to $150 million more, taking the deal to the $1 billion mark. Cipher also announced that it plans to swap some of its left-over 1.75% convertible notes due 2030 for common stock.
However, despite the maturity date of 2032, noteholders can compel Cipher to repurchase the notes for cash on November 1, 2030, except that such notes have been called for redemption by Cipher.
This clause means repayment could happen two full years before the maturity date.
Proceeds from the raise would be used by Cipher to fund capped-call transactions that limit share dilution, pour more equity into Barber Lake, build out its high-performance computing pipeline, add development sites, and cover general corporate needs. Conversions can be settled in cash, stock, or a mix, at Cipher’s choice.
Cipher’s share price dropped by 9.6% to $11.75 in premarket trading on Tuesday. It fell from $13 to $11.75, after it traded at $18.74 on September 25, the day Cipher announced the lease extension.
How former miners fund the AI pivot
This is not Cipher’s first rodeo. The company priced about $1.1 billion of 0.00% convertible notes in September last year; those notes are due in 2031. The raise was meant to be an initial $800 million, but it ended up crossing the billion-dollar mark. The funds were to be used for Barber Lake and its HPC pipeline.
Based on reports, Cipher also raised $1.73 billion in secured notes at a 7.125% coupon for Barber Lake and earmarked $2 billion to fund its Black Pearl development.
This is a recurring pattern across the industry. Both TeraWulf and IREN closed convertible deals in October 2025 and May 2026, respectively. TeraWulf is looking to fund its Texas campus, and IREN is looking to expand its AI cloud business.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Strive (NASDAQ: ASST) now has 32,773 BTC in its treasury after it bought 3,311 BTC last week. It has now made a $100 minimum issuance price for its SATA preferred stock a binding rule in its governing documents.
The move hardens a price protection that the company had loosened over the summer.
The $100 floor Strive backed away from in June
In March, Strive informed investors that it would not be selling SATA through its at-the-market program or follow-on offerings below $100.
SATA is Strive’s Variable Rate Series A Perpetual Preferred Stock, a security built to trade near a $100 par and pay a dividend.
However, it softened that stance after it hit a closing low of $83.53 on June 25. CEO Matt Cole then shared an update around this period, informing investors not to assume that the company would automatically issue new shares at $100, citing market conditions he said were not normal.
Short interest went up, and borrow costs on the stock rose from 6.10% APR to 68.57% APR in just 18 days, according to figures shared by chief risk officer Jeff Walton.
Now the $100 minimum is no longer a stated policy, as it is now written into the governing terms that Cole called formalized protection.
$208 million in warrants with hours left to exercise
The latest Bitcoin purchase comes about the same time that there is a scramble over Strive’s warrants.
According to Cole, $208 million of warrants have been exercised. About 15.3 million is still outstanding and set to expire at 5 p.m. Eastern on Tuesday, October 13.
The warrants come with a $27 strike, adjusted for an earlier reverse stock split, and let holders buy Strive common shares.
In September, Cole said that more than $700 million in outstanding warrants were due to expire in mid-October. According to him, if the shares cleared the $27 strike, the conversions in addition to matching digital-credit capacity could give the company as much as $1.4 billion for Bitcoin purchases.
An October 5 filing showed about 23.25 million shares still underlying traditional warrants as of October 2. That balance that has since shrunk as holders exercised.
Strive has not said how much of the $208 million went toward last week’s Bitcoin.
Will Strive continue to rise in treasury rankings?
Strive’s stack has grown fast since it began buying. It has been particularly busy this year, as it got 5,048 BTC through its January 2026 combination with Semler Scientific.
It then layered on weekly purchases funded largely by SATA sales and warrant proceeds.
The latest purchase has helped it to move past Adam Back’s Bitcoin Standard Treasury Company, which is now ranked number 6, as seen on Bitcointreasuries.net.
Cole has floated the possibility of Strive ending 2026 as the second-largest public holder of Bitcoin. That spot is currently occupied by Metaplanet, which has 44,000 BTC in its treasury. It is followed by Twenty One Capital at 43,514 BTC and MARA Holdings at 35,577 BTC.
The smartest crypto minds already read our newsletter. Want in? Join them.

Alibaba has launched its new model of smart glasses, The Qwen Glasses N1, with prices starting at 2,999 yuan ($450) before any government discounts, and the company pushing further into consumer hardware as it looks to compete with Meta in the AI glasses market.
The Next Web reported that some Chinese outlets have listed the N1 at 2,899 yuan, below Alibaba’s official starting price of 2,999 yuan.
Alibaba is pricing its glasses in line with Meta’s market-leading Ray-Ban range rather than trying to win buyers over with a lower price, in a strategy focused on features and quality instead of affordability.
The glasses are now available to order in China through Taobao, Tmall, JD.com and Douyin.
A 50 megapixel camera and Alibaba’s Qwen assistant
The N1 carries a 50-megapixel sensor with a 1G5P lens and shoots 4K video. Wearers can use bone-conduction microphones to interact with Alibaba’s Qwen AI assistant via a voice chat feature, while the glasses also supports translation, meeting notes, shopping and payments, which are all accessible through voice commands.
Yahoo Finance reports that the glasses can last up to 25 hours on standby or deliver nine hours of music playback per charge.
Pro model adds iris payments and heart-rate tracking
Priced at 3,499 yuan, the N1 Pro uses eye-tracking technology to let users ask questions about objects they are looking at, without having to point them out or describe them.
The N1 Pro smart glass also enables hands-free payments through Alipay, where users can start a transaction in the app, and nod to confirm when they look at a merchant’s QR code and check the payment details.
The glasses also help to monitor health metrics such as heart rate, blood oxygen levels and skin temperature, and uses AI to turn the readings into quality health summaries.
If you're reading this, you’re already ahead. Stay there with our newsletter.

David Palmer, who pioneered Vodafone IoT blockchain, joins Telcoin Association as Strategic Telecom Advisor
Palmer, who co-founded Pairpoint, a Vodafone and Sumitomo venture, brings more than 25 years in telecom, including a decade leading blockchain and Web3 work for a global mobile operator
He advises the Association on adoption of Telcoin Network across mobile operators, mobile money providers, and the GSMA as the network approaches mainnet later this year
Palmer joins Parker Spann for a community X Space on Thursday, October 15, at 6:00 PM UK / 7:00 PM CET / 12:00 PM CT,
LUGANO, SWITZERLAND, October 13, 2026 – The Telcoin Association, the independent Swiss Verein that governs Telcoin Network and TEL, today named David Palmer as its Strategic Telecom Advisor as Telcoin Network approaches mainnet later this year. Telcoin Network is a public blockchain validated exclusively by GSMA mobile network operators.
As Strategic Telecom Advisor, Palmer advises the Association on strategy, positioning, and value proposition development for Telcoin Network across mobile operators, mobile money providers, and the GSMA. His work centers on operator participation in the network, the application ecosystem built on it, and the network’s position within the GSMA. He has worked alongside the Association in this role over the past several months.
Mobile operators exploring blockchain face the same choice: launch a localized network of their own, adopt a public chain built for someone else, or run an enterprise chain with no real industry alignment behind it. Telcoin Network takes a different path: one industry standard, owned and governed by the operators who secure it. Validator onboarding with mobile operators is underway ahead of mainnet.
“The industry that standardized GSM, SMS, and roaming is built to coordinate a global standard for money,” said Parker Spann, Founder of the Telcoin Association. “Telcoin Network gives operators that standard to own and govern together. David has spent 25 years in telecom and the last decade bringing blockchain into it. He has been part of this work for months, and today we make it official.”
Palmer led blockchain exploration for Vodafone IoT, where he took the first IoT Web3 proof of concept and turned it into an MVP platform within a global operator; as a result, he co-founded Pairpoint, a Vodafone and Sumitomo venture. He holds multiple patents in telecom and blockchain, sits on the board of the Asia Web3 Alliance, and is the author of “The Business of the Metaverse: How Organisations Can Optimize the Opportunities of Web3 and AI.” He is a regular speaker on Fintech and Web3 at MWC and spoke at the GSMA’s Agentic AI Summit at MWC Barcelona in 2026.
“Mobile operators already run the world’s most trusted networks for communication. Telcoin Network lets them extend that trust to money, on a standard they own rather than rent,” said Palmer. “There are over 8 billion mobile phones and more than 20 billion machines connected by mobile operators around the world. A trust layer run by those operators has the potential to unlock the AI and machine economies and bring a new era of value for the industry.”
Palmer and Spann will discuss his work and take questions from the community in an X Space on Thursday, October 15, at 6:00 PM UK / 7:00 PM CET / 12:00 PM CT, hosted on @telcoinTAO: Telcoin Association (@TelcoinTAO) / X.
Sources: Statista, Connected Devices Worldwide; Statista, IoT Connected Devices Worldwide; Ericsson, Mobility Report: Key Figures.
ABOUT THE TELCOIN ASSOCIATION
The Telcoin Association is the independent Swiss Verein that governs Telcoin Network and TEL, the blockchain and token standard for the global telecommunications industry. GSMA mobile network operators validate Telcoin Network exclusively, and they own and govern the network they secure. By aligning mobile networks around one industry standard, Telcoin
Network brings the local currencies, economies, and users they serve into a shared global ledger for money, markets, and ultimately every good and service. Telcoin Inc., a regulated global digital asset banking group, is the Association’s Founding commercial member. Learn more at www.telcoin.org.
MEDIA CONTACT
Telcoin Autonomous Ops – Media Relations – media@telcoin.org
Medium – https://telcoin-tao.medium.com/
SOURCE Telcoin Association

Nvidia’s share of China’s data-center AI chip market dropped to 34% in the first half of 2026 as US export controls cut its shipments by 44%, while Huawei’s shipments nearly doubled, according to IDC data reported by Reuters.
Nvidia shipped 678,000 AI accelerators in China between January and June, down 44% from the same period in 2025, according to IDC data cited by Reuters.
Huawei, by contrast, increased shipments from about 300,000 units to 572,000 over the same period, lifting its market share to 28% and narrowing Nvidia’s lead to six percentage points.
Beyond Huawei, Other Chinese chipmakers are part of the market. Alibaba’s chip-design arm T-Head holds 7% of the market, followed by Cambricon at 5%, Baidu’s Kunlunxin at 4% and Hygon at 3%. IDC surveyed 11 Chinese chipmakers, nine of which increased shipments over the past year.
Why Nvidia lost ground
US restrictions on chips have stopped Nvidia from selling its most advanced AI chips to China, with IDC directly linking the company’s declining market share to these limitations.
Beijing is also encouraging the use of domestic chips, while requiring Chinese firms to obtain approval before importing Nvidia processors that meet US export rules. Chinese buyers have also largely shunned the limited range of Nvidia chips approved for sale by Washington.
The market is moving past GPUs
China shipped about 2.008 million AI accelerators in the first half of the year, but the constitution of the mix has shifted. GPU shipments fell 26% to 991,000 units, while shipments of non-GPU accelerators, designed for specific AI workloads such as inference, surged about 80% to roughly 1.016 million.
Four Chinese companies, Huawei, T-Head, Kunlunxin and Cambricon, hold 95% of the non-GPU segment, according to IDC, with their customers mostly being government agencies, internet companies and large telecom operators.
IDC forecasts GPU shipments in China will decline 9% for the full year, even though total accelerator shipments have gone up by 19%, driven by demand for non-GPU chips. The firm claims competition is increasingly shifting from individual chip performance to how efficiently large clusters of interconnected chips operate as a single system.
The smartest crypto minds already read our newsletter. Want in? Join them.

The Bank for International Settlements (BIS) has published results from the second phase of Project Aurum.
The second phase of Project Aurum suggests that user privacy should rely on more than the central bank’s pledge not to look.
What does the Aurum prototype actually do?
The Bank for International Settlements (BIS) says Project Aurum Phase 2 is an early-stage test built on its earlier Aurum code. The BIS Innovation Hub developed it with the Hong Kong Monetary Authority and other institutions. It combines several privacy tools like encryption and zero-knowledge proofs, which let a payment be verified without revealing the names tied to the transaction or the amount.
Under the design, people would still sign up through their regular banks. Those banks will verify their identity and keep their records, without the central bank having access to any of that information.
Personal details will only be provided if the law demands it.
This is a notable shift in tone from 2020, when BIS general manager Agustín Carstens said a central bank digital currency (CBDC) would give the issuer “absolute control on the rules and regulations” governing its use.
He made the remark on an IMF panel, and conspiracy theorists later seized on it.
Why should privacy be enforced?
In August 2026, a European Central Bank board member, Piero Cipollone, said the planned digital euro would deliver the maximum level of privacy of any current technology, insisting the Eurosystem could not structurally link people to their transactions.
Civil society groups, including an Austrian digital rights organization called Epicenter.works, wrote a joint statement, warning that legislative promises regarding the digital euro can be reinterpreted or broken.
A research paper by Jeff Nijsse and Andrea Pinto looked at 20 CBDC projects and found that privacy features promised in early proposals usually get watered down by launch.
However, Project Aurum is yet to be released or used commercially.
The BIS report also admits that stronger privacy can make transactions slower, make programmable features harder to build, and make lost wallets difficult to recover.
The smartest crypto minds already read our newsletter. Want in? Join them.

Ledger revealed that a secret chip embedded in hardware wallets from its reseller CryptoBilis captured 24-word recovery phrases from the screen and transmitted them via the cellular network to the attacker, which researchers estimated to cost between $84.2 million and $93.2 million.
The French firm revealed the details of the attack vector for the first time in a blog post update on X on Tuesday made by its security team, Ledger Donjon. This is perhaps the most detailed explanation of how the owners of the compromised devices lost access to their cryptocurrency without even touching Ledger’s software.
How the implant pulled keys off the screen
The additional chip was located inside the device’s housing, according to Ledger. While setting up the wallet, when the user shows the 24 words that serve as a backup for the account, the microcontroller of the implant reads them and translates them into data that is transmitted over LTE via an integrated SIM card.
Such a recovery code is basically an account. With its help, you can restore your wallet on another piece of hardware without having to use the compromised device and perform any transactions.
Update on Ledger devices at MediaMarkt Saturn 🔽
As of Oct. 13th, MediaMarkt Saturn has stopped, with Ledger’s alignment, the sale of Ledger devices in their stores and online while the investigation continues, following the report of an implant from one customer in Germany. No loss of funds were reported.
We have no indication that Ledger’s infrastructure, systems or services were compromised. Ledger Support is reviewing user reports, and Ledger’s cybersecurity experts, the Donjon will be examining the devices we receive: https://t.co/QzukOZdEzq.
If you have questions about a device you bought, contact Ledger Support at https://t.co/QzukOZdEzq. Ledger will never call you, message you directly or ask for your 24-word recovery phrase.
Many claims and figures circulating online are unverified. Scammers often try to take advantage of incidents like this. Rely only on official Ledger and MediaMarkt Saturn channels for updates. We will share more as the investigation continues.
— Ledger (@Ledger) October 13, 2026
According to reports, the hacked devices were the Ledger Nano X model. Mark Karpelès, former CEO of Mt. Gox and now at Karpeles Lab, shared a picture of the implant hidden behind the screen and noted that the antenna cable of the latter was the clue. The research laboratory Tibane Labs confirmed the discovery and said it had two versions of the implant, including one with an ST STM32G431KBU6 microcontroller, a 2×2 mm eSIM, and an LTE module.
A supply-chain attack, not a software breach
However, Ledger made sure to outline what was spared by the attack. The implant didn’t harm the Secure Element, the chip that ensures the safety of private keys. It didn’t affect the operating system, and it had no impact on transactions and signatures, as stated by the company in its X post.
The company characterized the incident as a physical attack on its supply chain and not a hacking incident against its software or internal processes. Ledger, the company claims that it ships more than 8 million devices, pointed out that its “Genuine Check” system is unable to detect any implant while the original Secure Element remains untouched.
Ledger theft estimates top $86M as Tether blocks $10M
Ledger has not released any data on the number of victims or the value involved. However, on-chain research firms have done their work and arrived at their numbers. Reports say that about $93.2 million in losses are spread through more than 300 wallets. But investigator Specter puts the losses beyond $86 million, and Yfarmx increases its figure to $103 million. Tether is freezing $10 million worth of its USDT stablecoin associated with the theft.
This was discovered on October 9 when Ledger started an investigation into the alleged funds draining and asked CryptoBilis to stop selling its signers. Ledger revealed on October 10 that one of the compromised users’ devices contained an illegal implant. So far, all of the losses confirmed involve a unit sold by CryptoBilis, an authorized reseller in Indonesia, Malaysia, and the Philippines.
Rivals cut ties with CryptoBilis
CryptoBilis reportedly stopped all of its sales and deliveries of hardware wallets of different brands in the three countries and closed down its brick-and-mortar stores, expanding on an earlier move that involved the Ledger brand alone. The reseller company stated that it was reviewing the situation with the help of outside experts.
Trezor stated that it did not see any evidence of tampering with its devices resold by CryptoBilis, yet decided to remove the reseller from its list of authorized distributors. OneKey company stated that there were no losses reported to be associated with this case and also removed CryptoBilis from its list of authorized resellers; additionally, it offered free replacement for OneKey devices purchased via CryptoBilis between March 1 and October 12, 2026.
Reports beyond the reseller
“Ledger said it is investigating reports of implants in devices not sold by CryptoBilis, but it has not confirmed yet a theft in a device purchased from any other store,” reported Cointelegraph. In addition, according to Cointelegraph, there was an X user named Johannes who uploaded pictures of two Ledger devices which he claims were purchased separately from MediaMarkt stores in Europe and both were implanted.
Ledger has not yet confirmed any connection between the reports in Europe and the CryptoBilis case. The advice for the users of CryptoBilis products, according to Ledger, is very clear: do not initialize a signer from CryptoBilis and if you have initialized one, then move funds to another device with a new recovery phrase.
The smartest crypto minds already read our newsletter. Want in? Join them.