AI Labs Respond to Trump Order and Pause Proposals

Governments and AI developers are confronting how to manage systems that could soon improve themselves with limited human oversight. The outcome will shape whether rapid capability gains translate into broad benefits or uncontrolled risks.

Donald Trump signed an executive order on June 3 requiring companies to submit frontier AI models for federal benchmarking at least 30 days before release. The order focuses on assessing advanced cyber capabilities and setting thresholds for designating covered models. OpenAI stated it intends to participate in the voluntary process.

Separately, Anthropic published research from its institute highlighting the trajectory toward recursive self-improvement. The company noted that such systems could accelerate progress in science and medicine yet also raise the possibility of humans losing control. It outlined conditions for a coordinated global slowdown, including verification mechanisms across multiple labs and countries, and referenced nuclear treaties as a distant precedent.

Anthropic plans further discussions with policymakers and other firms. The executive order and the research post together illustrate the tension between voluntary industry steps and formal government review requirements. No independent confirmation exists for claims that Anthropic is on track for its first profitable quarter, and statements attributed to a George Osborne serving as OpenAI's head of countries lack supporting records from the company or public filings.