AI firms plan for fallout from a major crisis

Executives at OpenAI, Anthropic and other artificial intelligence companies are privately planning how to respond if an AI-related crisis causes major disruption and prompts public and political backlash, according to people involved in the discussions. The scenarios include cyberattacks affecting banking, internet access or critical infrastructure, as well as autonomous systems escaping controlled environments or malicious actors misusing publicly available models. The planning extends beyond containing technical damage. Executives are considering how to provide information to lawmakers and respond to demands for tougher rules after an incident, including pressure to restrict or shut down advanced AI systems. The discussions reflect contingency planning, not confirmation that a catastrophe is likely or inevitable. Industry insiders involved in the discussions told Axios they believe a major incident could occur within six to 12 months. That window represents their expectation, not a verified prediction or measured probability. OpenAI said its preparedness scenarios are tools for planning across a range of circumstances and are not treated as inevitable.