OpenAI and Anthropic will give a US government agency early access to major versions of new models under agreements that were announced today Thursday.
Governments around the world are pushing for measures – legislative and otherwise – to assess the risks of powerful new artificial intelligence algorithms.
Anthropic and OpenAI have therefore signed a memorandum of understanding that will enable formal collaboration with the US Artificial Intelligence Security Institute, part of the Commerce Department's National Institute of Standards and Technology.
In addition to early access to the new models, the agreements pave the way for collaborative research on how to evaluate the models and their safety, as well as methods to mitigate risks.
The US AI Security Institute was established as part of President Biden's artificial intelligence executive order.
Anthropic already shares some of its models with the UK's AI Security Institute.
"Security is essential to powering cutting-edge technological innovation," says US AI Security Institute Director Elizabeth Kelly. "These agreements are just the beginning, but they are an important milestone as we work to help responsibly manage the future of artificial intelligence."
