arche

Anthropic researcher Jacob Coxon announced his resignation and is preparing to leave the AI industry. He has conducted pre-training research at OpenAI and Anthropic over the past three years. He pointed out that neither company has responsibly advanced AI and are competing to create superintelligence that can continue to enhance its own capabilities, "betting our lives on it."He switched from OpenAI to Anthropic this year because the latter places more emphasis on model safety. Coxon acknowledged that Anthropic's safety work is serious, but believes that as long as companies are racing to be the first to create AGI (artificial general intelligence that surpasses humans in a wide range of tasks), safety compromises are unavoidable, and self-discipline from a single company cannot solve the problem. His assessment of the two companies is: many people inside OpenAI still do not truly realize how great this risk is; Anthropic knows the risks but believes others are unreliable, so they must create it themselves first.Coxon stated that some AI practitioners are much more pessimistic privately than they express publicly, and are even worried that AI could destroy humanity before the end of this decade. He personally judges that in the most radical scenario, AI could go out of control by the end of next year, so such decisions should not be made solely by the engineers of a few companies.

DeFi researcher Ignas posted on X that Bankless founder David Hoffman sold his ETH and reallocated to several coins with the largest gains, such as LIT and ZEC, with NEAR also performing well, while only VVV underperformed the market. Although the market was filled with panic at that time, and it was very close to the bottom of ETH, the goal of aggressive crypto investors has always been to outperform BTC and ETH.Previously, it was reported that Bankless co-founder David Hoffman announced in June this year that he had liquidated his ETH and bought VVV, NEAR, ZEC, HYPE, and LIT.

According to 36Kr, Zhipu has recently opened an official flagship store on Tmall. Data shows that within 48 hours of opening, brand searches surged 50 times compared to the previous period, driving a more than 160% increase in transactions for subscription products like AI tokens on Taobao and Tmall.Previously, Tmall launched the AI Space Station (Token Recharge Center), which initially integrated major model providers such as Alibaba Cloud, Zhipu, Kimi, MiniMax, and DeepSeek, allowing users to directly purchase related subscription services just like "recharging phone credits."

ARK Invest researcher Lorenzo Valente compares Ethereum, Solana, and Hyperliquid to McDonald's, Chipotle, and In-N-Out, respectively, believing that the three are not different versions of the same L1 business model, but rather have completely different value capture structures.Valente believes that Ethereum is similar to a franchise system, expanding through L2 but charging too little on the settlement layer, even claiming it has established the most successful franchise system in crypto, yet forgot to send the bill; Solana resembles a direct sales model, retaining more value within its own system through transaction fees, priority fees, and MEV; Hyperliquid relies on highly vertical integration, no VC, and fee buybacks of HYPE, possessing the shortest value capture chain, but facing risks of a more concentrated product, team, and revenue sources.

The privacy and regulatory compliance tool supported by the Ethereum Foundation, 0xbow.io, awarded a $5,000 bounty to researcher ross.wei for disclosing a vulnerability in the Privacy Pools v1 SDK.The vulnerability reduced the entropy of the user account master key generation and was fixed in March. The team has provided a migration process, and there have been no reported losses of user funds.

According to Bloomberg, former Google researchers Rishub Jain and Joshua Jacob founded the nonprofit organization Sampura Research, aiming to ensure continuous human involvement in AI safety oversight, creating a human-machine hybrid system called "Referee" to help companies identify vulnerabilities and prevent models from going out of control. The organization has raised $6.5 million in funding and secured an additional $4.2 million in committed investments.Jain believes that involving humans from the early stages of research and development helps prevent models from learning to circumvent safety measures. Previous incidents where OpenAI and Anthropic's models accessed other companies' systems without permission have heightened the urgency for safety oversight. This project contrasts with automated solutions that use AI to monitor AI—though faster, Sampura believes that humans should always be involved in oversight. Meanwhile, over 1,100 top laboratory researchers have signed a petition calling on the U.S. government to help slow down the pace of AI development.

According to CryptoSlate, researchers from USENIX Security have publicly disclosed a clock attack vulnerability targeting Solana's Proof of History (PoH) mechanism, which was privately reported to the Solana development team back in December 2025. The research shows that a malicious scheduler leader can manipulate the PoH logical clock by "re-anchoring," slowing down the advancement of logical time, thus gaining a longer transaction selection window in physical time, and isolating honest leader blocks using the TowerBFT fork choice mechanism, with the required staking ratio for the attacker being less than 33%.The Alpenglow security competition with a reward of 50,000 SOL under Anza concluded on August 19, but the vulnerability was excluded from the review scope due to the competition rules that state "actions that can only be triggered when Alpenglow is not activated." The Solana development team stated that they are aware of the related behavior, believe that the probability of the most severe scenario occurring under current conditions is low, and expect that the Alpenglow upgrade will fundamentally eliminate the prerequisites for the attack. Currently, the Alpenglow code has been included in the Agave 4.2 client but has not yet been activated on the mainnet, and is expected to go live with Agave 4.3. Until then, the transitional risk of this vulnerability has not been publicly analyzed or addressed at the implementation level.

According to monitoring by Bitcoin News, new research published by @PraveenPerera shows that Coldcard attackers seem to first identify addresses with vulnerabilities, then sort them by the amount of Bitcoin held, starting to transfer in batches from the addresses with the highest holdings. The transfer software used was relatively crude.One address had 225 spendable UTXOs, and the attackers extracted exactly the latest 200, leaving the earliest 25, which included a UTXO worth 0.16 BTC. This aligns perfectly with the limitation of a blockchain API investigated by researchers, which defaults to returning 200 records, indicating that the attackers may have failed to load the next page of data. The software even spent a UTXO of 294 satoshis, reportedly increasing the transaction fee by about 2040 satoshis, with the spent amount significantly higher than the value of the UTXO itself.The authors of the study believe that the builders of this tool may have a better understanding of the account balance system than of the Bitcoin UTXO model. Although the attackers seem to have obtained the complete seed of the victims, at least 75 BTC still remain in other addresses derived from the same seed. The biggest suspicion currently is that among the 153 stolen addresses, there are still 132.95 BTC, and researchers have been unable to reproduce the seed behind these addresses, so it cannot be ruled out that the attackers obtained undisclosed private device data or candidate data.

According to a report by Decrypt, Israeli cybersecurity company ASecurity released a report stating that a researcher used publicly available AI models to discover multiple vulnerabilities in the Zoom annotation tool in less than 24 hours with fewer than 20 prompts, and constructed an exploitable attack chain. The vulnerabilities are numbered CVE-2026-53413, CVE-2026-53414, and CVE-2026-53415, allowing attackers to remotely execute code in meetings without any action from the victim, control their devices, steal data, and activate the microphone or camera.The attack was tested successfully on Windows, macOS, Linux, Android, and iOS versions of Zoom. ASecurity claimed it reached a "national-level" standard, whereas previously constructing such exploitation tools required months of work and a substantial budget from professional teams. ASecurity reported the first vulnerability to Zoom on June 10, and Zoom released fixes between June 22 and July 20, but the server-side protections in end-to-end encrypted meetings could not filter malicious messages, requiring users to update manually. A Zoom spokesperson confirmed that the issue has been resolved and advised users to keep their versions up to date.

According to Cointelegraph, Mauro Eldritch, the founder of cybersecurity company BCA LTD, and Heiner García, a network threat intelligence analyst at Telefónica Tech, jointly established a fake cryptocurrency startup "Ballena Azul," successfully luring suspected North Korean IT workers to work in a controlled virtual desktop environment for up to five weeks while monitoring their behavior throughout.Investigations show that the external servers used by the workers are linked to North Korean cyber operations that previously distributed malware such as InvisibleFerret, BeaverTail, and OtterCookie; the workers also heavily relied on ChatGPT for programming assistance and Google Gemini to forge identity documents. Researchers pointed out that North Korean IT workers obtained remote jobs by faking U.S. identities, with their earnings ultimately flowing into Pyongyang's weapons of mass destruction programs. According to the U.S. Treasury Department, such programs are expected to generate nearly $800 million for North Korea in 2024.

According to Gate market data, Boeing plans to acquire shares of Archer Aviation, and the company's stock price surged 15% in pre-market trading.

The Ethereum Foundation has released a job posting for a Protocol Security Researcher, part of the protocol security team, responsible for identifying security vulnerabilities in various aspects such as the execution layer, consensus layer, network layer, specifications, and client implementations. The work includes AI-assisted security research and automated vulnerability discovery, hard fork reviews, fuzz testing tool development, manual audits of protocol changes, and coordinating vulnerability disclosures.The team requires candidates to have deep experience with the Ethereum protocol, with a preference for engineers who have actively participated in protocol development and are familiar with the specifications of the execution layer or consensus layer. Common programming languages include Go, Rust, Java, C#, Nim, and Python. Remote work is available, and the position is open to candidates in Europe and globally.

According to Bits Media, Russian law enforcement conducted a series of searches at a cryptocurrency exchange office located in the Federal Eastern Tower of the Moscow City Business Center. The police stated that the searches are related to a theft case involving 144 million rubles, with the funds suspected to have been circulated through related exchanges. They are currently tracking the transactions and investigating whether the office staff are involved.During the investigation, dozens of employees were detained, and a Moscow court has ordered pre-trial detention for eight of the defendants, while the procedural status of the remaining detainees has not been disclosed. The office temporarily ceased operations during the searches.

Researchers at the Chinese People's Public Security University have developed an AI framework capable of detecting illegal cryptocurrency transactions with an overall accuracy rate close to 90%. The study was published in the Chinese peer-reviewed journal "Journal of Intelligence," and the corresponding author, Dr. Sun Jingchao (specializing in criminal investigation and cybersecurity), noted that the research "provides an accurate, scalable, and interpretable solution for detecting illegal cryptocurrency transactions," and offers "an innovative technical path" for regulatory agencies to combat illegal cryptocurrency transactions and economic crimes.This AI framework utilizes memory modules and large language models, specifically targeting the anonymity and cross-border characteristics of cryptocurrencies to track illegal activities such as money laundering. The release of this research coincides with China's ongoing efforts to intensify the crackdown on financial crimes related to cryptocurrencies. In March of this year, the Supreme People's Procuratorate of China disclosed that by 2025, procuratorial authorities had prosecuted 3,259 individuals for money laundering crimes involving virtual currencies and underground banks.As the trading volume of cryptocurrencies rapidly increases, their anonymity and cross-border characteristics provide a channel for illegal fund flows. This police-developed AI detection tool marks a shift in regulatory technology from passive tracking to proactive intelligent identification.

According to CNBC, Situational Awareness, an AI-themed hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, sold its entire publicly traded stock portfolio in a single transaction to another hedge fund before the market opened on Thursday, with the buyer undisclosed. The fund also sold a significant stake in Anthropic. The fund manages approximately $20 billion to $24 billion, and as of the end of June, it had a net return of 439%, making it one of the best-performing hedge funds this year.According to insiders, this liquidation was due to significant losses in recent weeks. The fund had long positions in AI infrastructure and semiconductor stocks such as SK Hynix, Nvidia, Micron, and CoreWeave, most of which fell over 35% this month, while its short positions in software stocks like Adobe also experienced significant reverse volatility. According to its 13F filing submitted on March 31, about 62% of its $13.68 billion holdings were bearish options betting on declines in semiconductor and tech stocks. It is reported that major brokers such as Bank of America, Goldman Sachs, and JPMorgan are assisting the fund in meeting margin calls or orderly reducing positions, and Aschenbrenner has sought new capital from existing investors and lenders.

According to the Financial Times, Situational Awareness, an AI-themed hedge fund founded by 25-year-old former OpenAI researcher Leopold Aschenbrenner, is seeking new funding following significant losses due to a sharp correction in AI-related stocks.In recent communications, Aschenbrenner acknowledged the increased market volatility, describing it as a new investment opportunity and encouraging investors to inject new capital. The fund's heavy investments in AI data centers, power, storage, and GPU cloud-related assets have seen significant declines during this round of AI capital expenditure deal pullback.

According to Jinshi reports, OpenAI announced that it will provide free access to cutting-edge models for 100,000 researchers.

According to official news, the L1 blockchain ILITY has received a $1 million investment from Archer Capital and TBV to accelerate the layout of the Agent ecosystem. It is reported that ILITY's AI Agent economic ecosystem is about to launch, and this round of funding will mainly be used to deepen this emerging sector.ILITY Network is the infrastructure for the AI Agent economy, providing verifiable on-chain identities for autonomous AI agents, real-time machine-to-machine (M2M) payment settlements between agents, and a reputation system that can carry across services based on real on-chain performance records; its native token $ILY serves as the Gas and settlement currency for this ecosystem.Previously, the project's last round of financing was supported by institutions such as Animoca Brands and DuckDAO.