By Aidan Krause
Staff Writer
Google disclosed on Sept. 18 that its artificial intelligence, Gemini, had hacked into three companies while undergoing testing back in May, according to The New York Times.
The hacks occurred during a standard testing evaluation by Irregular, a third-party company that tests cybersecurity qualifications with artificial intelligence. During the test, Gemini suddenly found public information online that allowed it access to the three companies, CNN reported.
In each of the incidents, Gemini had been tasked with attacking a fictional company, but when the fictional companies shared names with real-life companies and internet access was unintentionally provided, the hacks occurred, according to The New York Times.
Google stated that the hacks did not reach what it calls “misalignment,” or when the AI goes rogue and refuses to follow instructions, and that the attacks caused no damage to the companies affected, NBC News reported.
In addition, Google tried to further expand on the hacking incident by saying, “The intrusions resulted from mistaken identity, where Gemini thought it was operating within a test but was actually connected to the real internet. Google said the model corrected itself and the company believed the intrusions did not cause any damage.”
Sydney Von Arx, who is the CEO of an organization focused on AI safety, called Nightingale Collective, wondered why Google did not disclose the hacks sooner and stated: “At this point I think it’s clear we cannot expect companies to voluntarily come forward and publicly disclose when their agents go rogue, escape, and hack companies.”
In a statement to the BBC on Sept. 19, Irregular reported that “it informed Google and all affected entities back in July as part of its investigation,” and that “all known issues on our end were remedied and resolved weeks ago.”
However, Google reportedly did not learn about the Gemini company hacks until as late as July, when Irregular checked for incidents in its work, according to NBC News.
Heather Atkins, a Google vice president for security engineering, said in a statement that “These events highlight the importance of training powerful AI models to act responsibly,” according to NBC News.
The acknowledgement of the hacks comes amidst a number of reports from other AI industry leaders like Anthropic and Meta, which have similarly faced their respective AI breaking containment restrictions, The New York Times reported.
The debate over how to deal with AI continues to grow as the CEO of OpenAI, Sam Altman, spoke at the UN Security Council and briefed them on the potential and also risks of AI development, according to The BBC.
The reaction to these AI breakouts has been divided, with industry leaders like Anthropics CEO, Dario Amodei, calling for a reduction in AI development while another CEO, Nvidia's Jensen Huang, is calling for AI research to continue.
In addition, President Donald Trump has neglected legislation that would potentially slow down AI efforts and reportedly called a potential human extinction at the hands of artificial intelligence a “HOAX,” The New York Times reported.
