Labs work through crisis scenarios
According to Axios, Anthropic, OpenAI and other leading AI labs are working through scenarios in which a rogue AI wreaks havoc on critical infrastructure such as financial services, the internet, telecommunications, and power and water utilities. The leadership preparations are compared to the Pentagon's nuclear war games.
What the companies say
An OpenAI spokesperson said the company conducts preparedness exercises where teams discuss and work through a range of potential scenarios. The spokesperson said these scenarios are not treated as inevitable but are meant to help prepare for a variety of circumstances.
Risks and the regulation debate
The report notes that Anthropic listed "existential risks to humanity" as a major risk factor in its IPO prospectus, devoting nearly a third of the document to the issue. A former OpenAI safety employee argued that current risks stem from a lack of safety culture. The White House, meanwhile, has preferred to have companies self-police AI development rather than impose regulation.
Possible fallout
Axios says one focus of the scenarios is how to educate Congress after a major incident. Such an event would probably lead Congress to try to slow or shut down AI development, though planners think that would face significant challenges.
