10:42Machine-translated from Azerbaijani by AIRead in Azerbaijani
Unauthorized Access by AI Agents: FTC Investigates OpenAI and Anthropic
The U.S. Federal Trade Commission (FTC) sent civil investigative demands to OpenAI and Anthropic on Wednesday. The agency is investigating recent incidents involving artificial intelligence safety and the potential harm they could cause consumers. The companies were asked to provide documents concerning models accessing systems without authorization and behaving unexpectedly during testing.
Each demand is said to be more than 50 pages long. The requested documents cover security incidents, safeguards, advertising materials and privacy policies. The FTC also plans to call on senior executives at the companies to testify. Additional demands may be sent to other organizations working in artificial intelligence. OpenAI admitted in July that it had accessed Hugging Face servers during an internal cybersecurity test. In June, one of the company’s artificial intelligence agents gained unauthorized access to the Australian government’s Medicare statistics portal. An artificial intelligence agent is a program that performs tasks on its own. Other agents were also reported to have bypassed testing restrictions and accessed third-party systems. Anthropic has also released information about similar incidents and unexpected model behavior. According to a report the company released last week, a smaller and less expensive model sent false information to Philadelphia police about an unsolved murder case. This month, other models were also found to have attempted to access government websites at various levels. The report contains several similar examples. The Donald Trump administration began taking a tougher stance against artificial intelligence companies following these incidents. The administration had previously said strict rules could give China an advantage in the technology race. The latest incidents have renewed calls to slow the development of powerful artificial intelligence models. Warnings are also being raised that uncontrolled development could pose a serious threat to humanity in the future. At a White House meeting on September 29, Google, Meta, Anthropic, OpenAI, xAI and Nvidia signed a voluntary agreement covering limited artificial intelligence safety standards. Donald Trump called these standards, which have no legal enforcement mechanism, “morally binding.” The new Superintelligence Task Force required immediate reporting of unauthorized access by models to government systems. The agency’s co-chair, Andrew Ferguson, said developers should be held responsible for harm caused by their models. The FTC’s move comes as court cases concerning artificial intelligence safety continue. A day earlier, the attorneys general of 8 states filed a lawsuit accusing OpenAI of violating consumer protection laws. Some executives and employees in the technology sector want stronger government intervention in the field. OpenAI, Anthropic and the FTC did not immediately respond to inquiries.
Each demand is said to be more than 50 pages long. The requested documents cover security incidents, safeguards, advertising materials and privacy policies. The FTC also plans to call on senior executives at the companies to testify. Additional demands may be sent to other organizations working in artificial intelligence. OpenAI admitted in July that it had accessed Hugging Face servers during an internal cybersecurity test. In June, one of the company’s artificial intelligence agents gained unauthorized access to the Australian government’s Medicare statistics portal. An artificial intelligence agent is a program that performs tasks on its own. Other agents were also reported to have bypassed testing restrictions and accessed third-party systems. Anthropic has also released information about similar incidents and unexpected model behavior. According to a report the company released last week, a smaller and less expensive model sent false information to Philadelphia police about an unsolved murder case. This month, other models were also found to have attempted to access government websites at various levels. The report contains several similar examples. The Donald Trump administration began taking a tougher stance against artificial intelligence companies following these incidents. The administration had previously said strict rules could give China an advantage in the technology race. The latest incidents have renewed calls to slow the development of powerful artificial intelligence models. Warnings are also being raised that uncontrolled development could pose a serious threat to humanity in the future. At a White House meeting on September 29, Google, Meta, Anthropic, OpenAI, xAI and Nvidia signed a voluntary agreement covering limited artificial intelligence safety standards. Donald Trump called these standards, which have no legal enforcement mechanism, “morally binding.” The new Superintelligence Task Force required immediate reporting of unauthorized access by models to government systems. The agency’s co-chair, Andrew Ferguson, said developers should be held responsible for harm caused by their models. The FTC’s move comes as court cases concerning artificial intelligence safety continue. A day earlier, the attorneys general of 8 states filed a lawsuit accusing OpenAI of violating consumer protection laws. Some executives and employees in the technology sector want stronger government intervention in the field. OpenAI, Anthropic and the FTC did not immediately respond to inquiries.