US AI Safety Institute Partners With Anthropic and OpenAI

US Institute Partners with Leading AI Firms Anthropic and OpenAI on Safety

The US Artificial Intelligence Safety Institute has announced partnerships with AI leaders Anthropic and OpenAI. The collaborations aim to research and evaluate the safety of major new AI systems before and after their release to the public.

Access to Cutting-Edge Models

Through the partnerships, the Institute gains exclusive access to Anthropic's Claude 3.5 Sonnet and OpenAI's GPT-4. This early access enables comprehensive analysis of risks in these state-of-the-art natural language AI.

Providing Guidance on Improvements

A key aspect of the partnerships is the Institute providing feedback to both companies regarding potential safety issues and improvements for their AI models.

Collaborating with the UK

The US is also cooperating with the UK on standardized testing and auditing procedures to evaluate AI system safety before deployment.

Commitment to Transparency

OpenAI has expressed a commitment to transparency about the development process and safety considerations for their popular systems like GPT-4.

The US Artificial Intelligence Safety Institute's collaborations with major AI firms like Anthropic and OpenAI demonstrate an increased focus on responsible development of cutting-edge AI. Early intervention and guidance from experts will be key to building safe and trustworthy AI systems as the technology continues its rapid pace of advancement.