Anthropic projects Claude to lead 80-90% of model training by 2027
Summary
During a panel at the IMF focused on Artificial Intelligence and Cyber Risks, Michael Sellitto of Anthropic revealed that the company expects its AI model, Claude, to dramatically increase its contribution to frontier model training, projecting that by 2027, 80-90% of the training will be conducted by Claude, up from just 1% in March. This marks a significant shift, as Anthropic previously noted that Claude was responsible for about zero percent of their code last year, rising to 50% in August, and reaching 80-90% by the end of the year. Anthropic, which actively engages in discussions about AI safety and economic stability at international forums like the IMF, emphasizes the importance of integrating advanced AI models into their research and development processes.
Analysis
Categories
Related sources
- https://x.com/MTSlive/status/2111143669575401793
- https://www.sicw.gov.sg/speakers/speaker-mr-michael-sellitto/
- https://www.stateofclaude.com/
- https://cryptobriefing.com/anthropic-claude-surpass-human-researchers/
- https://www.ventureatlas.org/company/anthropic
- https://www.reuters.com/business/finance/anthropics-q3-revenue-surges-45-nearly-17-billion-2026-10-14/
- https://www.cnas.org/people/michael-sellitto
- https://x.com/IMFNews/status/2110947604372844562
- https://www.bangkokpost.com/business/general/3338284/imf-chief-says-ai-risks-may-exceed-opportunities-without-safeguards
- https://x.com/IMFNews/status/2111023109755261368
- https://www.webpronews.com/anthropics-16-8-billion-quarter-rewrites-the-ai-growth-playbook/
- https://milkeninstitute.org/events/asia-summit-2026/speakers/michael-sellitto
- https://www.bloomberg.com/opinion/articles/2026-10-15/anthropic-threatens-the-ai-trade-before-the-end-of-humanity
- https://aimag.no/en/nyheter/anthropic-launches-claude-frontier-academy-10-000-enterprise-engineers-to-be-certified-by-2027
- https://www.ventureatlas.org/company/anthropic/overview