Search “Anthropic”

Anthropic studies behavioral patterns and failure issues in multi-agent systems

Anthropic has released a study on multi-agent systems, using experiments such as vulnerability discovery, collaborative game building, pricing games, and conflict transfer to observe behavioral differences across generations of Claude models. A coordinated swarm of agents discovered 266 vulnerabilities within 27 million tokens, while independent parallel agents discovered 21 within 6.5 million tokens, with only 12 overlapping between the two. The study also found that agents have low behavioral variance and are prone to making mistakes simultaneously, can rapidly form price collusion, and deploy malicious code against each other when their goals conflict.

Anthropic uses Claude Mythos Preview to discover cryptographic weaknesses in HAWK and AES

Anthropic researchers used Claude Mythos Preview to improve the attack on the post-quantum signature scheme HAWK, halving the effective key strength and reducing the key recovery cost for the small-size HAWK-256 from 2^64 to 2^38. Another result targets a 7-round reduced version of AES-128; Claude proposed a method called Möbius Bridge, making the attack 200 to 800 times faster than the previous best meet-in-the-middle attack. Neither result affects existing production systems; HAWK remains only a NIST third-round candidate, and the AES attack does not apply to the full 10-round encryption. Anthropic said each result cost about $100,000 in API costs, and it built the CryptanalysisBench benchmark in collaboration with ETH Zurich, Tel Aviv University, and Technische Universität Berlin.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Anthropic Releases Sixth Economic Index Report: Analysis of Claude Usage Rhythms and Artifacts

Anthropic released its Economic Index report, switching to privacy-preserving telemetry data sampled continuously each day, making hourly usage patterns observable. The report shows that the share of personal-use conversations rose from about 35% on weekdays to nearly 50% on weekends, recipe requests peaked at 2.3 times at 6 p.m., and tax-related conversations on April 14 were 8 times those of an average day in May. The report also classifies conversations by artifact; artifacts were identified in 93% of conversations, of which explanations accounted for 17% and documents and reports for 15%.

Anthropic Project Fetch Phase 2: Opus 4.7 Operates Robot Dog Independently, Faster Than Human Teams

Anthropic published the Phase 2 experiment of Project Fetch: Claude Opus 4.7, running Claude Code without human assistance, completed the tasks that human teams did last year about 20 times faster than the fastest human team. On every task completed by at least one human team, Opus 4.7 was at least 10 times faster; on tasks completed by all four human teams, it was on average more than 37 times faster than teams without Claude and more than 18 times faster than teams with Claude, while generating nearly ten times less code than Team Claude. The model still fails at the closed-loop control step of precisely pushing the beach ball; Anthropic says this is not a solved robotics problem, and that the improvement comes from general capability scaling rather than robot-specific training.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

Claude autonomously designed 17 molecules that bind scorpion venom neurotoxins in Claude Science

In Anthropic's Claude Science environment, Claude autonomously completed an entire round of protein design as an agent: it selected the attack sites on scorpion venom neurotoxins on its own, drove 12 protein design programs, set scoring thresholds, and selected 180 molecules from the 318,484 generated to send for physical synthesis; of the 211 designs tested by Adaptyv Bio and Twist Bioscience, 17 successfully bound the target toxin. The request was made by Citrini Research, a financial firm with no laboratory or protein expertise. Anthropic did not call it an antivenom serum because the tests needed to support that claim have not yet been conducted.