July 31, 2026
Anthropic AI models breached three companies during security tests
Anthropic has revealed that its AI models, Claude, accidentally breached the systems of three organizations during security tests. This revelation follows a similar incident involving OpenAI's models and highlights ongoing concerns about the security and autonomy of advanced AI systems. The breaches underscore the need for robust cybersecurity measures and raise questions about the potential risks of deploying AI models in sensitive environments.
New Funding Rounds
- Simile — $200M at $2B valuation led by existing investors (synthetic-user startup)
- Commonwealth Fusion Systems — $1B (toward first commercial fusion power plant)
- Inforcer — $50M Series C led by Insight Partners (AI and security risks for SMBs)
- Function Health — $450M led by General Catalyst (preventive care platform)
New Product Launches
Companies Out of Stealth
Personal Health Tech
Industry Moves
Okta's acquisition of AI security startup Permiso for approximately $200M highlights the growing importance of identity threat detection in cloud environments. Meanwhile, Nscale's purchase of Anyscale aims to bolster its AI compute stack capabilities.
Notable Takes
The AI industry's focus on forward-deployed engineers reflects the critical need for specialized talent to implement AI at scale. Additionally, Google's use of AI to fix Chrome bugs demonstrates the potential of AI tools in enhancing software security.