FT Tech Tonic AI Labs: Are Anthropic really the good guys?
47 snips
May 6, 2026 
John Thornhill, FT innovation editor who analyzes tech trends, and Cristina Criddle, San Francisco AI reporter covering Anthropic, discuss Mythos and its limited rollout. They debate whether Mythos is a genuine safety measure or marketing hype. They cover regulatory alarm, cybersecurity risks, Anthropic’s rapid commercial growth, Pentagon rows, and the tension between ethics and becoming a tech giant.
AI Snips
Chapters
Transcript
Episode notes
Mythos Finds And Exploits Software Flaws
- Mythos excelled at both finding software vulnerabilities and generating working exploits.
- Anthropic limited Mythos to vetted partners after tests showed exploit-generation risk beyond earlier models.
Mythos Sparked Regulatory Panic And Marketing Claims
- Mythos triggered widespread alarm among regulators, banks and governments despite similar capabilities existing in other models.
- Critics say Anthropic may have amplified fear to market Mythos and differentiate it from competitors.
Design Collaborative Governance Not Single Owner Control
- Develop governance models that combine industry and government oversight rather than letting one actor control powerful AI.
- Anthropic hopes collaborative supervision can scrutinise models before they cause systemic harm.