Executives at Anthropic, OpenAI and other AI companies are privately gaming out scenarios for the political and public revolt that would follow a catastrophic AI-driven event, Axios reported on 9 October. The internal planning, described as "the day after" work, proceeds from a shared conviction among many leading researchers and company insiders that a serious incident is not a question of if but when.
The companies anticipate a large-scale cyberattack capable of shutting down financial services, internet connectivity, or even power and water supplies. According to Axios, many AI industry insiders believe such an event could occur within the next six to twelve months.
The scenario planning goes beyond standard crisis management. Axios reported that the work involves red-teaming worst-case situations and, critically, racing to brief members of Congress before regulators move. Executives reportedly understand that new AI legislation has no prospect of passing today, but they want to pre-position themselves to shape whatever laws US leaders reach for in the immediate aftermath of a disaster.
That political calculation is complicated by the November midterm elections. Axios reported that planners assume Democrats, expected to make gains, will move quickly to restrict AI, with some of the most aggressive proposals including banning superintelligence or pausing advanced AI development altogether. A bipartisan kill-switch requirement has also been floated, though experts have questioned whether turning off all AI systems is practically viable.
The concern is not purely hypothetical. A recent campaign targeting South Korean financial organisations including reported breaches at two banks illustrated the scale of harm a single actor can now inflict. According to CrowdStrike, a China-based suspect allegedly used an open-source AI penetration-testing tool called ARTEX, powered largely by DeepSeek, alongside Anthropic's Claude Code, to steal data from tens of thousands of bank customers. CrowdStrike senior vice-president Adam Meyers said the attack demonstrated how AI agents allow one person to target many victims in a very short time.
OpenAI told Axios that it conducts preparedness exercises covering a range of potential scenarios but does not treat them as inevitable. Anthropic declined to comment.
Source : Axios
