What happened
Anthropic is running an internal AI model called Model 2, described in the company's August 2026 Risk Report, that outperforms every publicly available version of Claude. On Anthropic's internal capability index (AECI), Model 2 scores about 1.5 points above Claude Mythos 5, though this gain is smaller than the jump from Mythos Preview to Mythos 5.
Why it matters
The model is deployed internally for coding, data generation, research and engineering tasks—sometimes through autonomous agents running continuously. Claude now writes most of the code in Anthropic's production systems, meaning the company's core operations depend on this unreleased capability. Anthropic found no new misalignments during internal review and rates the overall risk from misalignment as low.
What to watch
There are no plans to release Model 2 externally right now. The model was not tested as thoroughly as Claude Mythos 5 before deployment, and it is slightly weaker than Mythos 5 in some areas, suggesting development priorities remain focused on internal use rather than public availability.
Summaries like this, in your inbox every morning.
Anthropic's August 2026 Risk Report reveals that the company has been operating with internal AI capabilities that significantly exceed its public offerings. Model 2 represents an incremental but meaningful advance over Claude Mythos 5, gaining 1.5 points on the company's internal AECI benchmark—a narrower margin than the prior generation jump from Mythos Preview to Mythos 5. This suggests that capability scaling, while continuing, is showing signs of slowing down relative to earlier transitions.
The intensive internal deployment of Model 2 for code generation, data creation, and autonomous research tasks underscores how central AI has become to Anthropic's own operations. The fact that Claude now writes most production code across Anthropic's systems indicates the company has fundamentally restructured its development workflows around AI assistance. This reliance on an unreleased model for core business functions carries both operational advantages and risk management implications—advantages in speed and scale, but concentrated dependency on a system that underwent less rigorous testing than the public Mythos 5.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
In Q2, billionaires or their funds sold six 'Magnificent Seven' AI stocks
Nvidia was rising early Friday, even as the wider chip sector kept falling, Barron's reported
ElevenLabs completed a $300 million employee tender at a $22 billion valuation, and co-founder Mati Staniszews…
On October 15, OpenAI announced the Rosalind Academic Plan, giving approved US university labs free access to…
Neostellar Capital principal Evan Schlossman said CoreWeave's $103.7 billion in remaining performance obligati…
A Qiita article compared automated red-teaming methods for LLM apps — GCG, PAIR, and TAP — using mock implemen…
