Insiders expect a large-scale attack within 6 months to a year, though some security pros say the worst-case warnings are overblown.
Nothing to worry about over here, just the fact that the people building AI are preparing for it to go badly wrong, very soon. Axios reports that top executives at Anthropic, OpenAI and other AI companies are privately planning for a public and political backlash after a catastrophic AI event. Many insiders expect this to happen within six to 12 months, most likely as a cyberattack that knocks out financial services, internet access or even power and water.
Anthropic CEO Dario Amodei raised the alarm last month, warning that a swarm of AI agents could take over the internet, “potentially causing hundreds of billions of dollars in damage.” Some security pros pushed back. Expel CTO Greg Notch called the idea “far-fetched.”
Real examples are happening, though. A hacker allegedly used Chinese models, including DeepSeek, to steal data from tens of thousands of South Korean bank customers, then used Anthropic’s Claude Code to look for buyers, according to CrowdStrike.
The companies’ planning focuses on educating Congress. OpenAI says its preparedness exercises cover many scenarios. “These scenarios are not treated as inevitable,” a spokesperson told Axios. Anthropic declined to comment.
Nothing to worry about over here, just the fact that the people building AI are preparing for it to go badly wrong, very soon. Axios reports that top executives at Anthropic, OpenAI and other AI companies are privately planning for a public and political backlash after a catastrophic AI event. Many insiders expect this to happen within six to 12 months, most likely as a cyberattack that knocks out financial services, internet access or even power and water.
Anthropic CEO Dario Amodei raised the alarm last month, warning that a swarm of AI agents could take over the internet, “potentially causing hundreds of billions of dollars in damage.” Some security pros pushed back. Expel CTO Greg Notch called the idea “far-fetched.”
Real examples are happening, though. A hacker allegedly used Chinese models, including DeepSeek, to steal data from tens of thousands of South Korean bank customers, then used Anthropic’s Claude Code to look for buyers, according to CrowdStrike.
The companies’ planning focuses on educating Congress. OpenAI says its preparedness exercises cover many scenarios. “These scenarios are not treated as inevitable,” a spokesperson told Axios. Anthropic declined to comment.
