{"rewrite":{"id":"r_0e24820c98e8d263c41e703f","clusterId":"c_1421c51112551df7852486fd","slug":"anthropic-says-26-of-its-r-d-work-now-reaches-level-4-ai-autonomy","model":"deepseek-v4-1-flash","headline":"Anthropic Says 26% Of Its R\u0026D Work Now Reaches Level 4 AI Autonomy","summary":"Anthropic published internal metrics on its own AI use. As of August 2026, 26% of its research and development work had reached Level 4 on its R\u0026D Automation Index, where AI carries out most work under broad human instructions. Adding Level 3, the figure exceeds 90%. About 30,000 AI agents ran simultaneously on its main internal agent platform. No measured task reached Level 5, fully autonomous.","whyItMatters":"A frontier lab is now publishing its own internal automation numbers, which lets outside evaluators check the pace claims that labs otherwise make about AI developing AI.","webCardHtml":"\u003cp\u003eThe index classifies AI involvement across six levels, from Level 0, where humans do all the work, to Level 5, where AI runs from discovering a task to executing it. Anthropic based the index on evaluation criteria proposed by the research organization Epoch AI. Level 4 means AI carries out most of the work given broad human instructions.\u003c/p\u003e\u003cp\u003eOn monitoring, Anthropic combines checks on agent operations before execution with analysis of activity logs afterward. In August 2026 it monitored more than 1 billion decisions, and staff review about 50 high-priority items each week. For the week of July 13 to 20, 2026, about 6% of research and development compute went to safety-related work, rising to about 12% when limited to AI doing AI research.\u003c/p\u003e","blueskyPost":"Anthropic's own numbers show 26% of R\u0026D at Level 4, with about 30,000 agents running at once. Nothing measured hit Level 5, so the frontier is instruction-following at scale, not self-direction.","twitterPost":"Anthropic's metrics say 26% of R\u0026D hit Level 4 and 30,000 agents ran at once. No task reached Level 5.","threadsPost":"Anthropic reports 26% of its R\u0026D at Level 4 and roughly 30,000 AI agents running at once on its internal platform. Adding Level 3 pushes the figure past 90%. No measured task reached Level 5, so the ceiling Anthropic can document is delegation, not autonomy.","newsletterBlurb":"Anthropic published a set of internal metrics on how much of its own research and development AI handles. As of August 2026, 26% of that work reached Level 4 on its R\u0026D Automation Index, and Level 3 and above passed 90%. The company says it will keep publishing the numbers and will accept third-party evaluators.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260918-anthropic-measuring-claude/\",\"title\":\"About 30,000 AI agents run simultaneously inside Anthropic, with Claude leading 26% of AI research and development\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4639,"outputTokens":644,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1789709374,"createdAt":"2026-09-18T05:25:33.000Z","publishedAt":"2026-09-18T05:28:26.000Z","updatedAt":"2026-09-18T05:28:26.000Z"},"cluster":{"id":"c_1421c51112551df7852486fd","canonicalTitle":"Anthropic社内で約3万体のAIエージェントが同時稼働、AI研究開発の26％をClaudeが主導","representativeArticleId":"a_b7bfdcced9d0820303d252c1","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Claude\"],\"studios\":[\"Anthropic\"],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-09-18T04:30:00.000Z","lastSeenAt":"2026-09-18T04:30:00.000Z","updatedAt":"2026-09-18T05:28:21.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260918-anthropic-measuring-claude/","title":"Anthropic社内で約3万体のAIエージェントが同時稼働、AI研究開発の26％をClaudeが主導"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Claude"],"studios":["Anthropic"],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":null}