Today’s AI systems are remarkable, exciting, and offer the promise of substantial benefits to society. They also create risks, some of which are well understood and others we are all learning about as the technology advances. Vigilance and adaptability are central to managing those risks.

Major AI developers have already agreed, on a voluntary basis, to publish safety and security protocols that describe how they assess, test for, and mitigate potential for severe risks associated with their models. Thanks in part to our work and the work of many others, they are now required to publish such protocols and to follow them if they operate in California, Illinois, or New York.

However, these are just transparency laws, and companies determine for themselves how they respond in the event of determining that their systems possess dangerous capabilities, as many of them now have. Moreover, none of these laws require a third party to assess whether an AI company’s work presents a danger to the public, or task a government agency with the responsibility and authority to intervene if and when such a danger exists. We are advocating for state and federal legislatures to take these steps.

Our work is funded by individual donors and nonprofit institutions who believe in our mission. We do not accept corporate funding or funds from foreign governments.

If you are interested in partnering with us or want to learn more about our work, please visit our contact page.