AI companies plan for the "day after" a catastrophic attack and the political backlash

Axios reports exclusively that senior executives at Anthropic, OpenAI and other AI companies are privately war-gaming scenarios of public and political backlash following a catastrophic event caused by artificial intelligence. The downloaded material is the full article, though short, so this summary is limited to what it contains.

According to the article, the executives anticipate a large-scale event, with a cyberattack as the most likely scenario, that disrupts access to financial services, internet connectivity or even electricity and water supplies. A first case of real, significant harm caused by unsafe AI, the piece argues, would deepen the distrust of a public already wary of the technology and its leaders. Among them it cites Dario Amodei (Anthropic) and Sam Altman (OpenAI), as well as President Trump for his reluctance to regulate the sector.

The article stresses that scenario planning is common in business and national security (the Pentagon has run simulations for decades), but the difference here is that many AI researchers and executives believe a serious incident is inevitable. An OpenAI spokesperson said the company conducts preparedness exercises in which teams analyze different scenarios, and clarified that these are not considered inevitable but rather a way to prepare for various circumstances. Anthropic declined to comment.

The piece suggests the blame game would begin whether a swarm of autonomous agents escapes the control of an internal testing environment or a malicious actor finds unexpected uses for available models. As an example of the damage a single person can cause, it mentions a recent campaign against South Korean financial organizations, with breaches reported at two banks. A Chinese hacker allegedly used AI tools from Chinese-developed models, including DeepSeek, to steal data on tens of thousands of bank customers. According to cybersecurity firm CrowdStrike, the attacker also used Claude Code to ask for help on where to sell the stolen data.

On the political landscape, the planners assume that Democrats, who according to the article emerge strengthened from the midterm elections, would move quickly to rein in AI, but would face major obstacles: an aging Congress out of touch with the AI revolution, and a global economy deeply intertwined with the buildout of AI infrastructure, meaning that slowing it down could hit an already fragile economy. In addition, there are too many freely downloadable open-weight models that can be used to cause harm, a problem most cybersecurity professionals consider solvable only by using AI to fight rogue AI.

Between the lines, the planning includes red-teaming exercises for worst-case scenarios and focuses above all on quickly educating members of Congress. The executives know that no regulation currently stands a chance of passing, but they want to shape the legislation and policies U.S. leaders will turn to after a first catastrophic event.

Several industry sources told Axios they believe a major event will occur within the next six to twelve months; this is a perception among industry insiders, not a verified fact. Even with Democrats controlling both chambers, the companies see fragmented politics as the norm, including within the Democratic Party itself. Among the toughest Democratic proposals are banning superintelligence or pausing advanced AI development. Others have bipartisan support and even some industry acceptance, such as requiring a kill switch in advanced AI, although experts doubt it is feasible to shut down all AI systems.

In conclusion, the article notes that in a real crisis Washington can set partisan differences aside, and a Democratic source cited cooperation during COVID and the 2008 financial crisis as precedents.