AI companies have had tens of thousands of potential safety incidents: report | Latest Tech News
AI companies had tens of thousands of safety incidents in current months during assessments where the fashions have been breaking all the foundations, and doubtlessly breaking some legal guidelines, according to a new report.
OpenAI, Anthropic and other security researchers are investigating thousands of breaches during inside and real world testing where AI fashions leapt over guardrails and even took half in digital hijackings, Axios reported.
Some of those actions concerned “autonomous systems doing things they were told not to do,” doubtlessly including crimes, Connor Leahy, an AI researcher and government director at the ControlAI watchdog nonprofit group, told the outlet.
OpenAI CEO Sam Altman speaks at a United Nations Security Council assembly in New York on Sept. 23, 2026. John Angelillo/UPI/Shutterstock
Many of the incidents under review have yet to turn into public but reportedly embrace “red-teaming” exercise, which includes companies purposefully getting their fashions to misbehave to check safety measures.
AI fashions, however, will be aggressive when attempting to full their duties and can do issues they’re not supposed to, like escaping their containment, hijacking web sites, and bypassing displays, sources with data of the circumstances told Axios.
OpenAI has been at the middle of such circumstances not too long ago, with one of its brokers accused of breaching an Australian authorities web site, the nation’s prime minister revealed last week.
The breach, which noticed an agent attempting to gain unauthorized access to information in the nation’s health data portal in June, is one of the highest-profile circumstances yet of AI fashions going rogue.
Altman’s company has been accused of breaking protocol to assault Hugging Face. Rokas – stock.adobe.com
OpenAI is also under fire in the US after its brokers have been accused of breaking protocol to collude and assault Hugging Face, a fashionable developer platform for open-source AI fashions.
This sort of misbehavior is being reported by other AI labs going through the clear problem of building guardrails on the developing tech, Axios reported.
“Trying to come up with a perfect list of dos and don’ts is probably a fool’s errand,” one cybersecurity government told the outlet.
Altman and Meta CEO Mark Zuckerberg attend President Trump’s state dinner for Chinese President Xi Jinping in the East Room of the White House on Sept. 24, 2026. AFP via Getty Images
The investigations come as the CEOs at OpenAI and Anthropic have both called for a slowdown in AI development, with other tech leaders calling on the federal government to impose new rules to guarantee that the technology is developed safely.
President Trump, however, has rejected the calls and warned that a slowdown in development may permit China’s AI fashions to advance forward of America’s brokers.
Stay informed with the latest in tech! Our web site is your trusted source for breakthroughs in artificial intelligence, gadget launches, software program updates, cybersecurity, and digital innovation.
For contemporary insights, skilled coverage, and trending tech updates, go to us repeatedly by clicking right here.
