The cover is a GPTOnline editorial illustration accompanying this account of the October 2023 Preparedness announcement, not a photograph of the team or its work.
OpenAI, the trailblazing organization behind the creation of ChatGPT, announced in October 2023 that it was stepping up its commitment to artificial intelligence (AI) safety. The firm unveiled a new initiative dedicated to proactively addressing the full spectrum of potential risks associated with AI.
Dubbed Preparedness, this novel division was announced on October 26, 2023 and marks a significant advance in OpenAI‘s ongoing efforts to ensure the safe deployment of AI. The team will concentrate on identifying and mitigating potential AI threats spanning chemical, biological, radiological, and nuclear domains, as well as risks related to individualized persuasion, cybersecurity, and autonomous replication and adaptation.
The Preparedness team, announced under the leadership of Aleksander Madry, will grapple with critical questions such as the potential misuse of frontier AI systems and the threat posed by malicious actors deploying stolen AI model weights.
OpenAI described frontier models as potentially beneficial but capable of posing increasingly severe risks. That was the company’s rationale for preparing to assess and mitigate catastrophic risks, not evidence that any particular feared outcome was inevitable.
The announcement described recruitment for the Preparedness team and an AI Preparedness Challenge to identify catastrophic misuse risks. OpenAI offered $25,000 in API credits to each of up to 10 top submissions. The announcement page now marks the challenge as completed and lists its winners; this is not an open call for entries.
Preparedness was distinct from the Superalignment team announced in July 2023, which focused on steering and controlling much more capable systems. The October announcement focused on evaluating frontier-model risks and developing a Risk-Informed Development Policy.
As AI continues to evolve, concerns around its potential risks have grown. Some fear that AI could surpass human intelligence, leading to unforeseen consequences. Despite these concerns, companies like OpenAI continue to innovate in the field of AI, which has sparked further debate.
In May 2023, the nonprofit organization Center for AI Safety published a statement on AI extinction risk, urging the global community to prioritize mitigating the risks of extinction from AI alongside other societal-scale risks such as pandemics and nuclear war.
