OpenAI, Microsoft, Google, and Anthropic Launch New AI Security Group

OpenAI, Microsoft, Google, and Anthropic have collectively introduced the launch of the Frontier Mannequin Discussion board, an AI security analysis group devoted to making sure “secure and accountable growth of frontier AI fashions.” The brand new trade physique will work on figuring out and sharing greatest practices and advancing AI analysis.

“The Discussion board defines frontier fashions as large-scale machine-learning fashions that exceed the capabilities at present current in essentially the most superior current fashions, and might carry out all kinds of duties,” the 4 firms mentioned of their joint press launch.

The brand new Frontier Mannequin Discussion board can be opened to all organizations creating these “frontier fashions” in a secure and accountable approach. It would additionally collaborate with governments, policymakers, teachers, and civil society on AI security efforts.

“Corporations creating AI expertise have a duty to make sure that it’s secure, safe, and stays below human management,” mentioned Brad Smith, Vice Chair & President of Microsoft. “This initiative is a crucial step to deliver the tech sector collectively in advancing AI responsibly and tackling the challenges in order that it advantages all of humanity.”

At present’s announcement comes simply a few days after seven high AI firms (together with the 4 concerned within the creation of this new Frontier Mannequin Discussion board) made a dedication to the White Home to develop AI in a accountable approach. Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI all promised to have their AI programs examined by impartial specialists and collaborate on a watermarking system for AI-generated content material.

We’re at a wierd second the place generative AI expertise has turn into a key component of the competitors between the main tech firms together with Microsoft, Google, and Amazon. But, these firms additionally acknowledge that they should cooperate to make this groundbreaking expertise secure.

Final week, OpenAI unexpectedly gave us a very good instance of why these tech giants have to cooperate on AI security: OpenAI has simply discontinued its AI Classifier, which was designed to differentiate AI-generated textual content from human-written textual content. The corporate mentioned from the start that its classifier was not totally dependable, however it determined to finish the experimental undertaking on account of its “low fee of accuracy.”