Anthropic relaxes AI safety commitments, sparking a new competitive landscape with OpenAI

On February 25, news reports indicated that Anthropic recently adjusted its AI safety policy, removing the previous commitment to halt training advanced AI systems until safety measures are fully in place. This change gives Anthropic greater flexibility in competing with OpenAI, Google, and xAI, while also sparking industry discussions about balancing rapid AI development with safety assurances.

Anthropic’s Chief Scientist Jared Kaplan stated that stopping training doesn’t truly solve safety issues, and in the face of competitors’ rapid progress, unilateral suspension of development is unwise. The company emphasized increasing transparency by releasing cutting-edge safety roadmaps and regular risk reports, and delaying development when significant disaster risks are identified.

Senior policy researcher Edward Geist from RAND Corporation pointed out that early AI safety concepts originated from a specific group concerned with powerful AI, but the current large language models (LLMs) have redefined this idea. The policy wording changes send signals to investors and policymakers: companies are seeking risk management while maintaining economic competitiveness.

OpenAI also revised its mission statement in 2024, removing the word “safety” and framing its goal as “ensuring that artificial general intelligence benefits all of humanity.” These adjustments demonstrate that industry leaders are shifting toward more flexible and scalable strategies in both policies and messaging.

Recently, Anthropic completed a $30 billion funding round, valuing the company at approximately $380 billion; OpenAI is also raising up to $100 billion. Against this backdrop, AI companies face geopolitical competition and considerations for government contracts. The U.S. Department of Defense’s continued collaboration with Anthropic remains uncertain, leading to a more cautious and strategic policy approach.

Overall, the policy adjustments by Anthropic and OpenAI mark a new phase in the AI industry’s balancing act between rapid development and safety regulation, reflecting corporate strategic flexibility under capital influence and political pressure.