Ethereum co-founder and Consensys founder and CEO Joseph Lubin said in a post that, based on the investigation so far, there is no indication that MetaMask wallets or customer funds held in them were affected by a MetaMask infrastructure incident. As a precaution, the team and its partners rotated validator keys, meaning the affected validators must first exit the staking queue and then re-enter it.

Lubin said validator keys and withdrawal keys are separate and that his team does not hold clients' withdrawal keys, which he said prevents the incident from causing unauthorized transfers of staked ETH. The disclosure points to a security issue reaching part of Consensys' staking infrastructure.

According to Lubin, mnemonic phrases, private keys and wallet assets were not involved and remain self-custodied by users. He said the team has not publicly discussed details of the incident that is still being handled, but has disclosed it to partners and stakeholders and is working on the issue with them.

NEAR co-founder Illia Polosukhin (@ilblackdragon) said on X on October 3 that funds stolen from NEAR Intents had been fully recovered by 14:30 UTC that day, less than 24 hours after the attack. He credited the AI security layer SHIELD on Intents, along with active investigative work, with identifying the responsible party and opening communication with them.

Full restitution of exploited funds is uncommon in crypto hacks, so recovery within a day is a materially different outcome for affected users than the usual total loss. Polosukhin said NEAR will keep strengthening its security systems and urged the industry to respond with the same urgency and to cooperate.

Polosukhin said AI may be accelerating exploits, but it can also help defend against them, and that NEAR will continue adapting its security systems to a changing cybersecurity landscape. He framed the episode as a case for industry-wide vigilance and cooperation.

Aleph Alpha released Kolibri, an open-weight large language model optimized for German and English, together with a technical report that documents its dataset and training recipe. The report is published at aleph-alpha.com/downloads/tech-report.pdf.

It is an open-weights release from a European sovereign-AI lab, and commenters singled out the technical report's unusually detailed disclosure of dataset construction and training as a notable degree of openness.

The accompanying technical report is presented as a step-by-step account of how a modern agentic LLM was built, including how the dataset was assembled.

Chainalysis published a report on Wednesday attributing the $387 million Bitget exchange hack of Sept. 24 to actors tied to the Democratic People's Republic of Korea, saying the heist pushed the total value of crypto stolen by North Korea-linked groups in 2026 past $1 billion. The firm said it had been working with Bitget and law enforcement to trace the funds across multiple blockchains since the attack.

The attribution adds the weight of a major blockchain analytics firm to a growing consensus: Bitget CEO Gracy Chen said the attack's patterns matched North Korean hackers, and Elliptic called a DPRK link "highly likely." Chainalysis also said the incident pushed North Korea-linked crypto theft in 2026 past $1 billion.

In the first three hours, $387 million left Bitget across 23 transfers, landing on Ethereum (49.7%), XRP (40.8%), Zcash (7.6%) and Tron (1.8%). Chainalysis said in-house AI automation compressed an estimated 20-plus hours of manual bridge reconciliation into under 10 minutes, while stressing the technology accelerated investigators rather than replacing them.

Aave Labs has proposed creating the Aave Foundation, a memberless Cayman Islands foundation company that would hold legal title to the Aave trademark and related protocol intellectual property for the benefit of the Aave Protocol. The request, posted as an ARFC on the Aave governance forum and backed by founder Stani Kulechov, covers Phase 1 only: incorporating the entity and appointing its initial independent officers.

The proposal would give the DAO a recognized off-chain legal wrapper for its brand and codebase IP, a structural change for one of DeFi's largest lending protocols. According to the proposal, the foundation would not control protocol decisions, with AAVE token holders retaining existing governance powers.

Formation is described as the first of several phases, with trademark and domain transfers to be subject to separate governance votes. The entity would be a memberless Cayman Islands foundation company.

BNY Mellon (BNY) is in talks with Kraken parent company Payward over a potential digital-asset infrastructure partnership, people familiar with the matter said, in a report dated October 3. The discussions could cover digital assets, custody, trading, payments, wealth management and other financial market infrastructure, with services to be offered to banks, exchanges and asset managers through Payward Services.

A tie-up would connect the world's largest custodian bank with the infrastructure that has powered Kraken, extending BNY's existing custody, clearing, payments and wealth-management businesses onto digital-asset rails. The talks remain unconfirmed and are described as early-stage discussions sourced to unnamed people.

The arrangement could resemble the infrastructure component of Payward's recent commercial agreement with Nasdaq, according to CoinDesk. BNY already provides custody, asset servicing, clearing, payments and wealth-management services to institutional clients, while Payward Services says it gives fintechs, banks, exchanges and onchain platforms direct access to the infrastructure behind Kraken.

David Robinson, who wrote the safety reports accompanying OpenAI's major model releases, resigned this week and is publicly warning about AI safety risks in an editorial published by The Atlantic. According to reports on his departure, he spent about three and a half years at OpenAI, led the drafting of its current Preparedness Framework, and oversaw safety reporting on 12 frontier launches.

Robinson authored the safety documentation that accompanied OpenAI's major model releases, so his public criticism adds to outside scrutiny of how frontier labs conduct and disclose safety oversight. Reports on his exit say he wants AI labs to draw lessons from aviation and nuclear safety practices.

Robinson ran OpenAI's safety transparency work and took part in developing the system cards that describe its models, according to reports. One report headline described him as calling OpenAI's culture "broken," a characterization attributed to him in press coverage.

AI startup Jev has reached a $10 billion valuation, and founder Diogo Almeida has answered a wide range of questions about the company and its direction. The Q&A follows Jev's emergence from stealth as a model built by TypeSafe.

A $10 billion valuation, as reported, would place Jev among the most highly valued young AI companies, underscoring how quickly investor attention has concentrated on new model entrants.

Jev is the model developed by TypeSafe, the San Francisco startup founded in 2024 by Diogo Almeida, Erik Gafni and Sasha Sheng, and its name references the Jevons paradox. TypeSafe describes Jev as a frontier-intelligence function call — unstructured input in, typed probabilistic decisions out — rather than a conversational chatbot.

Anthropic announced the Claude Frontier Academy, a $100 million commitment to train 10,000 Frontier Deployed Engineers by the end of 2027. The program is described as a residency rather than a course, in which organizations nominate their strongest engineers for hands-on work with Claude.

The program ties Anthropic's Claude ecosystem to enterprise engineering teams at scale, with the stated goal of bringing participants to the standard of Anthropic's own deployed engineers.

According to the program description, participating organizations nominate engineers who build real Claude projects inside their own organizations; one account describes a multi-day in-person segment followed by a 12-week build period.