BTC $84,734 +1.3%ETH $2,702 +0.5%Fear & Greed 74 Greed

Today at a glance

Crypto's custody and solvency risks surfaced at once, while AI frontier competition shifted toward agents and security.

3 signals
  • Exchange solvencyBitMart published a preliminary proposal for its balance-sheet gap, offering immediate pro-rata distribution, a recovery token or a continuation token, pending court approval.#01
  • AI agentsOpenAI unveiled Dots at DevDay, an agent powered by GPT-6 Astra and positioned against Meta's freely distributed Muse platform.#02
  • Security lossesCertiK data put Q3 crypto security losses at $1.26 billion, with the $387.5 million Bitget hack accounting for about 31%.#03

Stories are ranked by impact; the first three are the edition highlights. This edition displays 14 of 420 candidates.

#01
CryptoEdition highlightTracking · 6 updates
9.0

BitMart Publishes Preliminary Asset Distribution Proposal, Plans Three User Exit Options

BitMart released an initial intent proposal and roadmap to address its operating difficulties and balance-sheet gap, converting outstanding user account balances into USD amounts to be independently verified by a court-appointed party, with three options: a front-end pro-rata distribution of fiat, stablecoins and major tokens for immediate exit; a recovery token (one unit per USD of balance) tied to clawing back stolen assets; or a DEX-tradable continuation token backed by investment rights and illiquid asset realisation. BitMart said it will consult its top 50 users by account balance over the next three to four weeks and is preparing a report on expected recovery ratios for each option to share with all users.

The proposal covers how all user balances are converted and settled, so affected users' access to funds is placed under a process that BitMart says requires court approval and court supervision rather than ordinary withdrawal. Foresight News and BlockBeats both describe the plan as an initial intent, not an approved distribution.

According to BlockBeats, BitMart attributes the gap to the December 2021 hack, which it values at about $319.5 million based on Dec. 4, 2021 prices, plus losses from the 2026 market downturn and from wash-trading groups that it says profited improperly from rebate and zero-slippage incentive policies. Under the continuation-token option, holders would participate in investment rights and non-current asset disposals and would share in distributable future profits only if funds are sufficient to restart the business.

telegram · foresightnews · · 2 sources

Background, discussion, and references

Market impact

BitMart is a mid-tier centralized exchange, so direct exposure is concentrated in its own users and balances; the transmission channel is counterparty and custody risk sentiment, which can prompt scrutiny of other venues carrying unresolved hack-related liabilities or large balance-sheet gaps. If the DEX-tradable continuation token is issued, it would create a separate secondary-market instrument whose value would depend on recovery outcomes under the court-supervised process.

Background

BitMart is a digital asset trading platform operated by bachi.tech in the Cayman Islands, known for a December 2021 security breach in which hackers used stolen private keys to take assets from hot wallets. HackenProof's post-incident bounty listing puts the December 2021 loss at roughly $196 million, while BitMart's proposal values the resulting reserve loss and balance-sheet gap at about $319.5 million using Dec. 4, 2021 prices.

References

Tags

#BitMart#exchange-insolvency#asset-distribution#recovery-token#court-supervision

#02
AI & TechEdition highlightTracking · 3 updates
8.5

OpenAI’s new agent is a shot at Meta — but can it compete with free?

At OpenAI's annual DevDay conference, CEO Sam Altman announced Dots, an agent product powered by the newly unveiled GPT-6 Astra model, describing it as a "real-deal AI" agent. The launch is explicitly positioned against Meta's Muse agent platform, whose early growth The Verge characterizes as runaway success.

According to The New York Times, Dots represents OpenAI's most aggressive push to build a product out of AI agents, which autonomously perform tasks for users. The Verge frames the open question around whether OpenAI's paid-tier agent can gain traction against Muse, which Meta distributes as a free download for Mac and mobile.

OpenAI describes Dots as always-on agents and says specialist dots are being brought to Microsoft Agent 365. GPT-6 Astra was released to the general public on September 4, 2026, with GPT-6 Sol and GPT-6 Luna following on September 22, 2026, according to Wikipedia; OpenAI says Astra is state-of-the-art on computer use and browsing, and it scores 59.3% on the Agents' Last Exam benchmark.

rss · The Verge AI · · Single source

Background, discussion, and references

Market impact

There is no direct product link to crypto, but frontier-lab agent launches shape narrative flow in the AI-agent token segment, which is the nearest point of market exposure. The competitive framing against Meta's free Muse also feeds the broader debate over pricing and distribution for agent products, an argument decentralized-agent projects commonly use for positioning.

Background

Meta announced Muse, its personal AI agent for long-running tasks beyond single-exchange chat, on September 8, 2026, and offers it free for Mac and mobile. OpenAI's GPT-6 family arrived shortly before, with Astra going public on September 4, 2026. The DevDay launch of Dots sets the two agent platforms up as direct competitors.

References

Tags

#OpenAI#GPT-6 Astra#AI agents#Meta#Muse#DevDay

#03
CryptoEdition highlightTracking · 3 updates
8.0

Bitget’s $388M hack pushes Q3 crypto security losses past $1B

CertiK data shows crypto security losses reached $1.26 billion across 247 incidents in Q3 2026, up 53.9% from $819.4 million in Q2. The $387.5 million Bitget hack alone accounted for about 31% of the quarter's losses, making it the largest single incident recorded under CertiK's methodology.

The figures show that a single exchange breach accounted for nearly a third of the quarter's industry-wide losses, with incident counts rising about 13% from 219 to 247 quarter-over-quarter.

CertiK recorded roughly $769 million in losses across 99 incidents in September, of which about $273 million was frozen or returned, leaving adjusted losses of $495.3 million; exploits accounted for $734 million, or nearly 96% of the month's losses across 58 incidents. Bitget detected unauthorized transfers from some hot wallets on Sept. 24 and suspended withdrawals, saying attackers exploited a vulnerability in a third-party security product to obtain internal credentials and forge withdrawal commands.

rss · Cointelegraph · · Single source

Background, discussion, and references

Market impact

The quarter's losses cluster in centralized exchange hot-wallet custody, Bitcoin-sidechain token backing and on-chain lending: Bitget's Sept. 24 withdrawal suspension, the Liquid Network exploit involving unbacked L-BTC, and the Tectonic lending exploit on Cronos all fell into those segments. Aggregate loss figures at this scale feed into how exchanges, bridges and DeFi protocols assess custody, third-party vendor risk and reserves.

Background

CertiK is a blockchain security firm that tracks and audits Web3 protocols. The quarter's second-largest incident was a $319 million exploit on the Liquid Network, a federated Bitcoin sidechain built by Blockstream, on Sept. 6. Tectonic followed at $120 million, and the Coldcard theft at $112.7 million. SlowMist traced Bitget hack activity to an Aug. 31 zero-day exploit, predating the publication of the quarterly report.

References

Tags

#Bitget#exchange hack#crypto security#CertiK#Liquid Network#Tectonic

#04
Crypto
8.0

Drift Foundation Launches DFX Claims and Redemption, 1 DFX Per 1 USDT of Loss

Drift Foundation announced on October 2 that DFX claims and redemption are live, allowing users with verified losses from the April 1 incident to claim 1 DFX for every 1 USDT of loss. Each DFX carries a claim on a Recovery Pool that currently holds roughly 3.11 million USDT.

The launch turns the April 1 losses into a tradable, pool-backed claim structure, giving verified victims a defined redemption path rather than an open-ended recovery promise. The pool's size relative to total losses means the redemption amount per DFX depends on how much is ultimately recovered and committed.

DFX is redeemable for USDT at a Redemption Amount calculated as the Recovery Pool balance divided by DFX outstanding supply; redeemed DFX is burned and the transaction is irreversible. The Recovery Pool is funded by a daily share of Velocity's protocol net revenue, up to 127.5 million USDT in matching funds from Tether, up to 20 million USDT from strategic partners, and assets recovered later; the claims window closes at 00:00 UTC on January 1, 2028, after which unclaimed DFX is burned.

telegram · theblockbeats · · Single source

Background, discussion, and references

Market impact

The mechanism ties Drift's recovery directly to USDT: Tether's commitment of up to 127.5 million USDT in matching funds and the partner commitments would flow into the Recovery Pool, while the value of DFX is determined by pool balance against outstanding supply. Exposure concentrates in Drift's user base and the Solana DeFi segment, and the 2028 claim deadline sets the timeframe over which recovery proceeds can reach the pool.

Background

The April 1 security incident drained roughly 295.4 million USDT in user assets, according to the recovery update Drift Foundation published the day before the claims launch. The foundation has engaged Mandiant, zeroShadow and SEAL 911 for the investigation and fund tracing, and Mandiant identified the attacker as UNC6862, a North Korean threat group.

References

Tags

#drift-protocol#exploit-recovery#defi#claims-redemption#north-korea-hacking#tether

#05
8.0

MetaMask unstakes $1.4B of ETH after validator rewards stolen

MetaMask proactively initiated the exit of about 17,000 affected Ethereum validators holding more than 523,000 staked ETH (roughly $1.4 billion) after block rewards from 18 of its validators were found to have been diverted to an address funded through Tornado Cash. The company said its wallets faced "no immediate threat" and that the issue was confined to its "non-custodial staking operations".

The mass exit pushed Ethereum's validator exit queue to its highest level this year, with the estimated withdrawal wait extending from about three and a half days to nearly two weeks. MetaMask has not disclosed whether validator signing keys were exposed, leaving slashing risk and the full scope of the breach unresolved.

Security researcher 0xKaden, of Spearbit and Cantina, said block rewards from 18 MetaMask validators were "not paid to the correct fee recipient but instead to this tornado [Cash] funded account"; the address has made no transactions, holds 0.46 ETH including 0.1 ETH funded from Tornado Cash, and received 18 incoming ETH transfers from Titan Relay: Forwarder worth less than $1,000 in total. On-chain analyst Emmett Gallic separately flagged a 133,300 ETH transfer (about $360 million) from wallets labelled "Lubin/ConSensys" a few hours before MetaMask's statement, with no suggestion that transfer was suspicious.

rss · Protos · · 5 sources

Background, discussion, and references

Market impact

The exits move roughly 523,000 ETH through Ethereum's staking withdrawal pipeline, swelling the exit queue to 773,447 ETH — its largest backlog since December 2025 — and delaying when those funds become liquid, while Lido expects the ETH to be re-deposited after an exit, withdrawal and re-entry cycle estimated at up to 45 days. The channel of exposure runs through staking infrastructure and liquid-staking supply rather than spot custody, so further disclosure on signing-key exposure or slashing would shape sentiment toward non-custodial staking providers.

Background

Tornado Cash is an open-source, non-custodial privacy protocol on Ethereum that mixes transactions to obscure their origin and destination, and it has been sanctioned by the U.S. Treasury. The incident resembles a prior case in which Kiln suffered a $41 million loss in its SOL staking operations and responded by exiting all active ETH validators and rotating signing keys, treating related operations as potentially compromised. In July, Drop Site News reported that ConsenSys, MetaMask's developer, accidentally hired a software developer linked to North Korea as a consultant for about a month, though that report does not suggest the role was related to the affected staking operations.

References

Tags

#metamask#ethereum-staking#security-incident#validators#tornado-cash

#06
AI & Tech
8.0

Google releases Gemini 4 Argon, called its most powerful model yet

Google released Gemini 4 Argon on Wednesday, September 30, 2026, marketing it as its most powerful model to date and pitching it as a workhorse for coding and cybersecurity work. According to CNBC, the model launched first with selected cybersecurity partners rather than general availability.

The release is Google's latest entry in the frontier-model competition, and per CNBC its initial rollout is limited to select cybersecurity partners rather than all developers. Google positions it for developers, professionals and enterprises tackling difficult problems.

Google's blog says Gemini 4 Argon was built with frontier-level capabilities in coding, knowledge work, cybersecurity defense and creative writing. Third-party tracker Artificial Analysis lists "Gemini 4 Argon (High)" among the leading models in intelligence and describes it as reasonably priced relative to similar models.

rss · TechCrunch AI · · Single source

Background, discussion, and references

Market impact

The release has no direct on-chain, custody or token-supply linkage; any crypto read-through runs through sentiment in AI-narrative assets such as decentralized-compute and AI-agent tokens, which have historically reacted to frontier-model announcements from major labs. Because availability is initially gated to select partners, no immediate developer-demand shift is established.

Background

Gemini 4 Argon uses a new naming scheme and follows the cancellation of Gemini 3.5 Pro, with Google having focused on Gemini 3.8 Flash, according to 9to5Google. Frontier models are typically trained on vast datasets at very high cost, covering data acquisition and processing plus compute. Third-party comparison coverage places Gemini 4 Argon alongside OpenAI's GPT-6 Astra as a same-generation rival.

References

Tags

#Google#Gemini 4 Argon#frontier model#AI release#coding#cybersecurity

#07
Policy
8.0

US Sanctions A7 Network, Proposes Ban on Sub-Agent Payments

The United States sanctioned the A7 Network, a Russia-linked payments operation, and Treasury's Financial Crimes Enforcement Network (FinCEN) proposed a rule that would prohibit transmittals of funds involving the network's Sub-Agents. FinCEN said A7's sub-agents processed more than $17 billion in dollar-denominated transactions from January 2025 through June 2026.

The proposed rule would extend US restrictions beyond A7 itself to the foreign sub-agents that appear as the contracting or paying party on A7-related invoices and payment instructions, a layer FinCEN says is used to move funds from non-Russian bank accounts. The payment prohibition is a proposal and is not yet in force.

According to FinCEN's notice of proposed rulemaking, the A7 Network assigns a foreign Sub-Agent to appear as the contracting or paying party on invoices, sales agreements and payment instructions, enabling payments to be executed from non-Russian bank accounts through correspondent banking and SWIFT channels. The measure was issued alongside Treasury action under the heading "Operation Economic Outcast."

rss · The Defiant · · Single source

Background, discussion, and references

Market impact

The proposed prohibition targets the sub-agent layer that routes payments out of non-Russian bank accounts, so the direct exposure sits with correspondent-banking counterparties, crypto exchanges and ruble-pegged stablecoin settlement rails linked to A7A5 or Grinex, transmitted through compliance and de-risking pressure rather than any change to protocol rules. A7-linked flows reported through global banks also point to heightened screening risk for institutions handling related counterparties.

Background

A7 is described as enabling cross-border money transfers to Russian individuals and entities, including those under sanctions, using bank infrastructure in Russia and Kyrgyzstan, the Kyrgyzstan-based crypto exchange Grinex (a successor to the sanctioned Garantex), and a network of shell companies in third countries. It issues the ruble-pegged stablecoin A7A5. An October 2025 report by the Centre for Information Resilience described A7's use of foreign shell companies and promissory notes as a form of trade-based money laundering, and said A7 claims over 30 shell companies in its network of "payer companies."

References

Tags

#sanctions#FinCEN#AML#A7 Network#A7A5#money-laundering

#08
8.0

SEC Proposal Would Address How Investment Advisers and Funds Can Custody Crypto Assets Under the Federal Securities Laws

The U.S. Securities and Exchange Commission proposed new rules and amendments establishing a tailored framework for the custody of crypto assets by registered investment advisers and regulated funds. The 760-page proposal, published Thursday and open for a 60-day public comment period, would clarify which entities may serve as custodians and how advisers and funds must keep records and make federal disclosures.

SEC Chairman Paul Atkins said the proposal "would provide a clear regulatory framework for the custody of crypto assets, giving investment advisers and funds a compliant pathway where none existed before." Atkins also said existing custody rules were designed to protect client assets from loss, theft, misuse and misappropriation, but considered only traditional assets — "an untenable situation in the 21st century."

The proposal would allow advisers to self-custody client assets in limited cases — a scenario an SEC official described as likely unusual — requiring that no qualified custodian is available, that the firm meet certain expertise requirements, and that the arrangement be reviewed quarterly to see whether a custodian has become available. It would also permit the use of state-chartered trusts as custodians, and includes new clarifications of industry practices and auditing requirements.

rss · SEC Press Releases · · 6 sources

Background, discussion, and references

Market impact

The proposal targets the custody channel through which institutional capital reaches crypto markets: it would define which entities — including state-chartered trusts — can hold client crypto assets, and set conditions under which registered advisers may hold them directly. Because it is a proposal subject to a 60-day comment period rather than a final rule, its effect on custody arrangements and asset-manager participation remains contingent on the rulemaking process.

Background

The proposal follows other recent SEC crypto-related releases, including last month's Innovation Exemption and August's Reg Crypto publications. It also follows prior SEC no-action relief allowing investment advisers to use state-chartered trust companies as qualified custodians for crypto assets. Commissioner Hester Peirce, who has led the agency's Crypto Task Force since its inception, leaves the agency on Friday; the SEC moved this week to reduce the number of commissioners required for a quorum from three to two.

References

Tags

#SEC#crypto-custody#regulation#investment-advisers#regulated-funds#state-trust-companies

#09
7.5

Strategy has burned through 64% of its STRC rescue fund

Strategy has spent $1.28 billion, or 64%, of its $2 billion Digital Credit Securities Repurchase Program on buying back STRC, leaving less than $724 million as of Monday's filing. STRC, which Strategy has said it wants to trade near its $100 stated amount, opened for trading at $99.18.

The fund meant to keep STRC near $100 is nearly depleted, while the shares have traded below $99 on the majority of trading days over the past year. The buybacks are funded largely by dilution of MSTR common stock, which ranks junior to STRC.

Strategy spent $151.7 million repurchasing STRC last week at an average price of $98.86 per share, up from an $86.52 average in late July, and has bought back 13.3 million STRC shares, roughly one in eight outstanding. Although the program technically allows repurchases of STRF, STRD and STRK, the source reports it has only ever repurchased STRC.

rss · Protos · · Single source

Background, discussion, and references

Market impact

Strategy is the largest corporate bitcoin holder, and its BTC purchases depend on capital raised through MSTR common and preferred stock; a shrinking repurchase cushion and continued common-share dilution affect the cost and capacity of that funding channel. That transmission runs primarily through MSTR and STRC as well as the wider bitcoin-treasury-company segment, and only indirectly to BTC demand-side flows.

Background

Saylor introduced STRC in July 2025 as a perpetual preferred stock paying a variable dividend, pitched as yielding income derived from BTC without BTC-style volatility. As BTC lagged the roughly 30% annual pace he had described, the board raised STRC's dividend to 12% and authorized a $1 billion buyback fund on June 29, then doubled that authorization to $2 billion by September 8. STRC traded below $72 per share as recently as June.

References

Tags

#Strategy#STRC#MSTR#preferred stock#Bitcoin treasury

#10
7.5

Open USD bets Stripe’s $1.9 trillion network can break USDT and USDC’s grip

Open USD (OUSD), a stablecoin issued for Open Standard, launched on Sept. 30 with 468.4 million tokens outstanding and about $468.45 million in backing assets. It is issued natively on Tempo, Base, Ethereum and Solana, and comes with a distribution partnership that makes it available across parts of Stripe's payments stack.

Stripe processed $1.9 trillion in total volume in 2025, and OUSD's current circulation is less than 0.3% of that monthly figure, illustrating the scale of the network through which the token could be distributed if businesses adopt it for payments and treasury activity. Stripe co-founder and CEO Patrick Collison said Open Standard was designed so most reserve yield flows back to participating partners rather than being retained by the issuer.

Reserve data showed $257.2 million in cash and $211.2 million in Treasuries and short-duration money-market funds, with more than $400 million of liquidity deployed across decentralized exchanges, stablecoin swaps and bridges on Tempo. Stripe said businesses can hold OUSD through Stripe Treasury, send it via Global Payouts, accept it with Payments and use it with stablecoin-backed card products, subject to product and geographic availability; Coinbase begins supporting the network Oct. 1, and additional Mastercard distribution through BVNK is planned.

rss · CryptoSlate · · Single source

Background, discussion, and references

Market impact

OUSD adds a new dollar-denominated venue competing for stablecoin float that currently concentrates in USDT and USDC, with distribution flowing through Stripe's payment rails and, from Oct. 1, Coinbase's exchange access, plus planned Mastercard reach via BVNK. The transmission channel is therefore payments and treasury demand plus exchange liquidity for a token whose supply is still about 0.15% of the sector, leaving the competitive effect dependent on whether launch liquidity converts into recurring usage.

Background

Dollar stablecoins are worth more than $300 billion, with Tether's USDT at roughly $183.8 billion and Circle's USDC near $74.1 billion, together accounting for about 84% of the market, according to CryptoSlate data. OUSD's roughly $468 million supply represents about 0.15% of the sector. Tempo, one of the chains where OUSD is natively issued, is described as a payments-first Layer 1 blockchain incubated by Stripe and Paradigm. Open Standard has assembled more than 200 financial institutions, fintechs, banks and businesses as partners, with Stripe, Visa, Mastercard, Coinbase and Shopify among its founders.

References

Tags

#stablecoins#OUSD#Stripe#Tempo#payments#market-structure

#11
7.5

⚡️NEAR Intents Attacker May Be Lazarus Group

According to on-chain monitoring by PeckShield dated October 1, the NEAR Intents attacker moved the stolen funds to KuCoin and bridged them into BTC. The attacker's address has interacted with an address (0x098B7...E2f96) flagged as North Korea's Lazarus Group.

If confirmed, the on-chain interaction would tie the roughly $3.8 million exploit to a group attributed to North Korea; NEAR Intents says the incident has been reported to law enforcement and that affected funds will be compensated in full.

NEAR Intents said the loss stemmed from a bug in how its Omni deposit and withdrawal infrastructure interacted with the NEAR Intents smart contract, and that the contract-side issue has been patched. The attribution to Lazarus Group has not been independently confirmed.

telegram · theblockbeats · · 2 sources

Background, discussion, and references

Market impact

The incident bears directly on the NEAR token and on the Bitwise NEAR ETF launched two days earlier, both of which declined after the exploit, while the routing of funds through KuCoin and into BTC touches exchange custody and cross-chain bridging as laundering channels. The disclosed loss of roughly $3.8 million is small relative to overall market liquidity, so any wider effect would travel through sentiment toward bridge and intent-based infrastructure rather than through systemic fund flows.

Background

NEAR Intents is a cross-chain swap service that routes orders through competing market makers and says it has handled more than $30 billion in swaps across 35 blockchains. Two days before the exploit it blocked a $50 million swap attempt from the Bitget hacker. Bitget CEO Gracy Chen and analytics firm Elliptic pointed to North Korea as the likely culprit behind the Bitget exchange hack, which has also not been confirmed. Cross-chain bridging layers have repeatedly been targeted, as in the 2022 Harmony bridge hack that cost about $100 million.

References

Tags

#NEAR Intents#Lazarus Group#exploit#money laundering#KuCoin#on-chain security

#12
7.5

Citi and Coinbase Expand Partnership to Bridge Fiat Banking and Stablecoin Payments

Citi and Coinbase have expanded their partnership to link conventional fiat banking infrastructure with stablecoin-based payment flows, according to The Daily Hodl. The move pushes deeper integration between traditional finance and crypto settlement rails.

As framed by the source, the expanded tie-up deepens the connection between a major global bank and stablecoin-based payment flows, extending institutional exposure to crypto settlement rails. No specific terms, products or launch timelines are given in the available reporting.

google_news · The Daily Hodl · · Single source

Background, discussion, and references

Market impact

The transmission runs through institutional stablecoin adoption and banking settlement channels rather than through token supply or on-chain liquidity: stablecoin issuers and payment-focused crypto infrastructure stand to gain a larger share of mainstream settlement volume if such bank partnerships scale, while Coinbase's listed equity is directly exposed given it is a party to the deal. Any effect depends on how far the integration is actually implemented.

Background

The announcement builds on an existing partnership between Citi and Coinbase. It comes as payment and crypto infrastructure firms race to connect bank fiat rails with stablecoin settlement; Fireblocks, for example, launched a "Network for Payments" with compliance tooling across more than 100 countries in September.

References

Tags

#stablecoins#Coinbase#Citi#institutional adoption#payments

#13
AI & Tech
7.5

OpenAI Says People Linked to China's Moonshot Tried to Copy Its AI's Hidden Reasoning

OpenAI says it disrupted a coordinated campaign to extract the hidden reasoning its models generate before answering, a campaign it says began on July 1. On July 24 and 25 alone, OpenAI logged 16,000 extraction requests from more than 4,000 users as part of a wider cluster of over 15,000 users, and it says the activity was fully disrupted by July 28.

According to OpenAI, the hidden reasoning it targets could be used to train another model without the original safeguards. OpenAI says it has since closed a pathway that let someone who already held another user's encrypted reasoning replay it to recover its contents.

OpenAI said the operators "did not break our encryption, compromise a database, or gain direct access to stored user conversations," but instead manipulated model interactions so protected reasoning could be reproduced in visible form, violating its terms of service; one method involved copying encrypted reasoning out of one conversation and asking a model to decode it in another. OpenAI also said it is unclear whether all operators observed originated from a single actor, while attributing a core cluster of the activity to individuals associated with Moonshot AI, developer of Kimi.

rss · Decrypt · · Single source

Background, discussion, and references

Market impact

The disclosure does not directly involve crypto assets; any transmission runs through sentiment, as renewed attention to model-output provenance and cross-border AI IP disputes can affect the AI-narrative segment of crypto markets, including AI-themed tokens and decentralized inference or compute networks. The channel is narrative and regulatory attention rather than cash flows or custody, and the report itself contains no market data.

Background

OpenAI has made similar accusations before: in January 2025 it said it was reviewing signs that DeepSeek may have distilled its models, as Washington weighed national security risks. Anthropic followed in February, accusing Chinese labs of using about 24,000 fraudulent accounts to generate more than 16 million exchanges with Claude; by April the White House was describing industrial-scale distillation campaigns by foreign entities, primarily in China, and in June Anthropic asked Congress for penalties on large-scale model extraction. In August, researchers showed that OpenAI, Anthropic and Google each protected reasoning with a single provider-wide encryption key, and that attackers could coax models into outputting the hidden thoughts in plain text; all three deployed server-side patches after disclosure.

References

Tags

#ai-safety#openai#moonshot-ai#model-distillation#security

#14
Crypto
7.0

Ethereum Foundation and Open Anonymity Launch zkAPI for Privacy-Preserving API P

The Ethereum Foundation and Open Anonymity Project launched zkAPI, a privacy-preserving payment layer for metered APIs, now live on Ethereum mainnet. Users deposit ETH, USDC or other credits into an onchain vault and authorize spending with zero-knowledge proofs, allowing providers to verify payment without learning the payer's identity or which deposit funded the usage.

According to Coinotag, the launch gives Ethereum's privacy stack a working mainnet showcase for zero-knowledge payments and gives the Ethereum Foundation a reference deployment it can point other teams to, while the AI inference side needs usage data before it can be called adopted.

zkAPI can issue short-lived, spending-capped API keys and settle based on actual usage, and it is initially focused on AI inference but can extend to other pay-per-use services. The system does not provide network or content anonymity, meaning IP addresses, timing patterns and identifying information in prompts may still be linkable.

telegram · wublockchainenglish · · Single source

Background, discussion, and references

Market impact

The primitive routes pay-per-use settlement through onchain vault deposits denominated in ETH and USDC, creating a new demand channel for vault contracts and privacy tooling on Ethereum mainnet; the extent of that effect depends on provider adoption and usage, which the available evidence does not yet establish.

Background

The launch follows earlier Ethereum developer discussion of a privacy model for AI chatbots that shields user identities while still letting providers verify payments, as reported by crypto-economy.com. Vault-based onchain deposit structures are an established pattern on Ethereum mainnet.

References

Tags

#ethereum#zero-knowledge-proofs#privacy#payments#ai-inference