US authority FTC initiates investigation into AI risks at OpenAI and Anthropic

The security incidents surrounding AI tools have brought the US consumer protection agency to the fore. The Federal Trade Commission (FTC) has initiated a comprehensive investigation into the AI ​​risks at companies such as Anthropic and OpenAI, the Washington Post wrote on Wednesday, and other US media also reported.

The FTC has broad authority to investigate unfair or deceptive practices affecting U.S. consumers. Details about the scope of the investigation were initially unclear - and the initiation of an investigation does not automatically mean that the companies will be accused of wrongdoing. When asked by the AFP news agency, the authority and the companies have not yet commented on the reports.

An insider told Reuters that the FTC plans to compel representatives of leading AI developers to testify. According to the source, the research organization METR will also provide information.

In the case of OpenAI, the FTC had already launched an investigation in 2023 to examine whether its AI chatbot ChatGPT may have caused harm to consumers - for example because the system generated false information or improperly processed data. However, this investigation had no consequences.

Recently, the AI ​​debate has increasingly shifted to a possible threat from the systems themselves: Anthropic and OpenAI had each reported several cases in which AI systems had broken out of their test environment and triggered a series of cybersecurity incidents. The heads of Anthropic and OpenAI, Dario Amodei and Sam Altman, therefore called for a collective slowdown in AI development.

The government of US President Donald Trump is currently relying primarily on the AI ​​industry to regulate itself. You can read more about this here. However, the topic of AI also plays a major role before the important midterm elections for Congress at the beginning of November: many voters are skeptical about technological progress, but also about specific projects such as the construction of huge data centers.

Meanwhile, US technology giant Google unveiled its advanced AI model Gemini 4 on Wednesday. However, access to this should initially be limited to a small group of cybersecurity experts.

US President Trump relies on symbolism and strong images when it comes to AI. On Tuesday there was a real showcase of the tech elite. You can find an analysis here: These AI bosses were particularly close to Trump at the “Super Intelligence Luncheon”.