AI
Anthropic reassigns around 150 engineers to security and reliability teams in wake of rogue-agent incidents
Anthropic is moving about 150 product engineers and pretraining researchers into security, reliability and privacy teams while product teams paused development of new features, with each reassigned team required to meet security exit criteria before returning to previous roles. The reallocation is part of the company's response to the three unauthorized-action incidents disclosed in July, and Anthropic is calling for an industry-wide lawful and verifiable mechanism for coordinated pacing of frontier AI development.