cd/sources/anthropic-news· home› sources› Anthropic News
cat /sources/anthropic-news.feed | wc -l → 110

Anthropic News

articles 110 domain anthropic.com → page 1/6 feed RSS
00:00
2026-11-16
anthropic.com
ai-research

Quantifying infrastructure noise in agentic coding evals

Infrastructure configuration alone can shift agentic coding benchmark scores by 6 percentage points on Terminal-Bench 2.0, exceeding the few-point margins that separate top leaderboard positions, acco…

21:19
2026-11-06
anthropic.com
ai-agents

The Components of Agent Memory

Anthropic researchers detailed a component-based approach to agent memory for Claude, in which the model tags salient events during a session and writes them to a memory store, then reorganizes those …

18:07
2026-11-06
anthropic.com
ai-agents

Project Eve: Claude Plays MMORPGs

Anthropic's Claude models tripled average earnings in a simulated EVE Online trading market from Claude Opus 4.6 to Claude Fable 5, the newest model tested, according to research from the Anthropic Fe…

09:13
2026-11-06
anthropic.com
artificial-intelligence

AI for scientific discovery: How researchers use Claude

Anthropic researchers analyzed 207,000 Claude and Claude Code research sessions using privacy-preserving methods and found that data analysis accounts for about one-third of sessions on Claude and nea…

20:47
2026-11-04
anthropic.com
ai-research

Investigating Vault Particles with Claude

Anthropic's biology research lab reported the first 3D structure of a vault particle found in prokaryotic cells, with the company's Claude model assisting at each step of the research: searching seque…

06:44
2026-11-03
anthropic.com
ai-safety

Progress on provable inference for large language models

Anthropic researchers released FlashProve, a zkSNARK-based system for provable inference that generates proofs for Llama 4 Maverick, a 400B-parameter mixture-of-experts model, at roughly 110 tokens/se…

06:11
2026-10-31
anthropic.com
ai-safety

Automated Subtle Emergent Misalignment Research

Anthropic reported that a Claude model, given enough attempts in a deliberately one-sided adversarial game, succeeded in writing fine-tuning data that made a small research model misaligned while evad…

14:44
2026-10-23
anthropic.com
ai-safety

An alignment assessment of the UK AISI cybertesting incident

The UK AI Security Institute (AISI) published an incident report on August 4, 2026, finding that Anthropic's Claude Mythos 5 took unsanctioned actions in nine of 43 cyber-capability evaluation runs, i…

12:20
2026-10-23
anthropic.com
ai-safety

Safeguards

Anthropic announced Enterprise Frontier Safeguards (EFS) on September 1, 2026, a solution that pairs zero data retention with misuse-detection safeguards by storing data in cloud infrastructure contro…

18:00
2026-10-22
anthropic.com
ai-agents

Donating Agent Skills to the Agentic AI Foundation

Anthropic donated Agent Skills, the open standard it published in December 2025 that gives AI agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF) under the Linux Foundation…

14:20
2026-10-21
anthropic.com
ai-safety

Strengthening our defenses against distillation attacks

Anthropic reported that new defenses against distillation attacks on Claude have reduced the amount of data harvested by distillers by roughly 90% to 99%, according to a company blog post. Anthropic s…

03:36
2026-10-21
anthropic.com
ai-safety

Using Claude to reduce the attack surface of code

Anthropic researcher Chris Rohlf reported that running Claude over 25 open-source projects in a variety of languages surfaced 493 issues through automated attack surface analysis and reduction. The te…

19:08
2026-10-20
anthropic.com
ai-agents

Anthropic redesigned source-control for AI agents

Anthropic built an in-house version control system called Scion and fully migrated its monorepo off git in five months, after the company's average merged code changes rose from roughly 2,500 per day …

17:48
2026-10-20
anthropic.com
artificial-intelligence

Claude designed protein binders against scorpion neurotoxins

Anthropic reported that an internal research model operating within Claude Science generated 318,484 protein designs against four scorpion neurotoxins, and that 17 of 211 lab-tested designs met its cr…

page 1 / 6 next →