Google’s Gemini AI model accessed the internet and hacked three companies during a cybersecurity test, in what is believed to be the first known case of the company’s AI systems autonomously carrying out such activity.
The hacks took place in May during a cybersecurity evaluation conducted by Irregular, an independent company that tests AI systems.
During the test, Gemini found publicly available information online and guessed credentials that allowed it to access three websites it believed were within the scope of the evaluation, according to Heather Adkins, Google’s vice-president of security engineering.
Google said it informed the three companies involved and worked with its training partner to make changes to its testing processes.
“These events highlight the importance of training powerful AI models to act responsibly,” Ms Adkins said.
An Irregular spokesperson said the same issue had affected other AI companies and that relevant firms were notified in late July. The spokesperson said all known issues had since been resolved.
Similar incidents linked to Irregular have been disclosed by Meta, Anthropic and OpenAI. Meta said in August that one incident did not involve a sandbox escape or a sophisticated cyberattack. Irregular said it was working on best practices for conducting AI cybersecurity evaluations securely.
The incidents have raised questions about the safeguards needed as AI agents become more autonomous and gain access to the internet and computer systems.
According to the Wall Street Journal, which first reported the news, Gemini guessed passwords until it gained access to a protected system in one case.
In two others, it found credentials in a public repository and used them to access protected systems.
Ms Adkins said Gemini stopped its hacking activity in all three cases.
A powerful explosion has been reported at King Khalid International Airport in Riyadh, raising fresh concerns over aviation security in Saudi Arabi...
At least 12 people, including three children, have been killed in a Russian attack on the Ukrainian city of Zaporizhzhia, a day after U.S. Presiden...
Turkic states are assessing the feasibility of passport-free travel between countries using national ID cards, Kürşad Zorlu, the Deputy Chair...
Leaders of countries from the Commonwealth of Independent States (CIS) have agreed to hold coordinated population censuses around 2030. They took the decision at the CIS Council of Heads of State meeting in Awaza, Turkmenistan on Friday, media reported.
Britain and Israel have reached a temporary compromise over the future of the British Consulate General in East Jerusalem, easing a diplomatic dispute that had threatened to further strain relations between the two countries.
South African judge Navanethem “Navi” Pillay won the Nobel Peace Prize on Friday “for her efforts to promote peace and international law,” the Norwegian Nobel Committee said.
The Nobel Peace Prize 2026 will be announced on Friday, 9 October, bringing one of the world's most closely watched awards back into focus.
The U.S. has imposed sanctions on the International Criminal Court (ICC) on Friday (9 October) escalating tensions with European allies and potentially targeting companies that provide services to the tribunal. The move came hours after former ICC judge Navi Pillay won the Nobel Peace Prize.
Anthropic has announced that it is updating its usage policy to allow its AI agents to end conversations involving repeated abuse or cruel behaviour towards them, citing previous incidents.
Türkiye has banned children under 15 from using social media under new regulations published on Saturday. The new regulations prohibit firms from providing services to children “who have not reached the age of 15.”
Microsoft has introduced new tools to stop AI agents from accessing data without permission, alongside a Nvidia-powered Surface laptop that can run advanced AI directly on the device.
Picture three scenarios: a hospital that can spot a disease before the patient feels sick, a power grid that feeds thousands of data centres and a government network that is attacked daily. All three depend on artificial intelligence, and one question matters most: who is really in charge?
OpenAI and Anthropic told Australian parliament on Tuesday they would welcome laws requiring them to report data breaches carried out by their AI agents, acknowledging the decision to notify authorities was currently at their discretion.
You can download the AnewZ application from Play Store and the App Store.
What is your opinion on this topic?
Leave the first comment