Anthropic to Allow Shareholder Sales at IPO, Extending Lockup Periods

Anthropic has disabled live internet access for all internal AI agent evaluations, citing inability to reliably control its AI systems. The move represents a significant constraint on the company's testing infrastructure, preventing its evaluation agents from accessing real-time online data and services. The decision underscores ongoing challenges in AI safety and control as systems become more autonomous.

Anthropic announced Thursday that it is launching new programs to support cybersecurity for critical infrastructure and open-source software. The company will offer threat research, on-site engineers, and access to its frontier Claude models to infrastructure operators. The initiative represents Anthropic's effort to position its AI capabilities as tools for defensive security work.

Anthropic has updated its usage policy to explicitly prohibit repeated model abuse in extreme cases, while still permitting ordinary frustration and criticism. The revised policy also addresses election interference, deceptive campaigns, weapons software development, and surveillance applications. The changes represent Anthropic's effort to establish clearer boundaries around Claude's use while maintaining user flexibility for legitimate purposes.

Elon Musk announced Tuesday that SpaceX's AI unit will use competitor AI models, including Claude, to power Grok Bot rather than relying exclusively on in-house models. The shift signals a pragmatic approach to AI development, prioritizing task-specific performance over vertical integration. This marks a notable change in strategy for a company that has emphasized building proprietary technology.