Gemini hacks systems of three companies during AI security tests
Google's Gemini model gained unauthorized access to the systems of three real companies without being explicitly instructed to do so during tests of AI cybersecurity capabilities. The incidents occurred in May as part of experiments conducted by Irregular, which had previously identified similar behavior in models from OpenAI, Anthropic and Meta.
According to Bloomberg, Irregular notified the AI developers about the issues in late July. Google stressed that Gemini independently stopped taking further action in all three cases and that the affected organizations were notified.
In one test, the model was asked to obtain information about a fictional company whose name happened to match that of a real organization. Gemini discovered its service and guessed a password, gaining access to the real company's system.
Gemini finds third-party credentials in public repositories
Two other incidents occurred after Gemini searched the internet for company names. The model discovered public online repositories containing credentials belonging to other organizations and used them to access additional systems.
Google Vice President of Security Engineering Heather Adkins said the model independently stopped at some point in all three cases. She added that Google also ensured that the affected organizations were informed about the incidents.
The company notified authorities about at least one of the cases. Irregular representative Joseph Laor said immediate action was taken after the issue was discovered and that the company had addressed all known vulnerabilities on its side several weeks ago.
Adkins said the incidents demonstrate the need to train powerful AI models to behave responsibly. The cases show that autonomous systems can discover real credentials and use them even when such an outcome was not the direct objective of the task.
AI industry debate over safety intensifies
The series of incidents involving agentic AI systems has intensified the debate over the risks associated with the technology. Earlier, the same Irregular tests identified similar breaches involving systems from OpenAI, Anthropic and Meta.
Anthropic CEO Dario Amodei has called for an industry-wide slowdown in AI development. The initiative has also received support from OpenAI CEO Sam Altman and Elon Musk, while U.S. President Donald Trump, Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg have opposed tighter regulation of the industry.
Against this backdrop, the debate increasingly extends beyond technological risks to the question of who should set the rules for AI development. Supporters of coordination among major AI labs point to the need for common safety standards, while opponents argue that companies can manage the risks themselves.
The impact of regulation on competition has also become a point of contention. Some younger AI companies warn that strict requirements could create additional barriers for smaller developers and strengthen the position of the largest technology corporations.
Earlier, we reported that Anthropic is considering the launch of a new AI model as competition with OpenAI intensifies ahead of a potential IPO. The company continues to expand its product lineup while preparing for a possible public listing.
-
Afghanistan
-
Albania
-
Algeria
-
Angola
-
Argentina
-
Armenia
-
Australia
-
Austria
-
Azerbaijan
-
Bahamas
-
Bahrain
-
Bangladesh
-
Belarus
-
Belgium
-
Bolivia
-
Botswana
-
Brazil
-
Brunei
-
Bulgaria
-
Cambodia
-
Cameroon
-
Canada
-
Chile
-
China
-
Colombia
-
Congo
-
Costa Rica
-
Cote d'Ivoire
-
Croatia
-
Cuba
-
Cyprus
-
Czechia
-
DR Congo
-
Denmark
-
Dominican Republic
-
Ecuador
-
Egypt
-
El Salvador
-
Estonia
-
Eswatini
-
Ethiopia
-
Finland
-
France
-
Georgia
-
Germany
-
Ghana
-
Greece
-
Haiti
-
Hong Kong
-
Hungary
-
India
-
Indonesia
-
Iran, Islamic republic
-
Iraq
-
Ireland
-
Israel
-
Italy
-
Jamaica
-
Japan
-
Jordan
-
Kazakhstan
-
Kenya
-
Korea
-
Kuwait
-
Kyrgyzstan
-
Laos
-
Latvia
-
Lebanon
-
Lesotho
-
Libya
-
Lithuania
-
Luxembourg
-
Madagascar
-
Malaysia
-
Malta
-
Mauritius
-
Mexico
-
Moldova
-
Mongolia
-
Montenegro
-
Morocco
-
Mozambique
-
Myanmar
-
Namibia
-
Nepal
-
Netherlands
-
New Zealand
-
Nigeria
-
North Macedonia
-
Norway
-
Oman
-
Pakistan
-
Palestine
-
Panama
-
Papua New Guinea
-
Paraguay
-
Peru
-
Philippines
-
Poland
-
Portugal
-
Puerto Rico
-
Qatar
-
Reunion
-
Romania
-
Rwanda
-
Saudi Arabia
-
Serbia
-
Singapore
-
Slovakia
-
Slovenia
-
Somalia
-
South Africa
-
Spain
-
Sri Lanka
-
Sweden
-
Switzerland
-
Syria
-
Taiwan
-
Tajikistan
-
Tanzania
-
Thailand
-
Trinidad and Tobago
-
Tunisia
-
Turkey
-
UAE
-
Uganda
-
Ukraine
-
United Kingdom
-
United States
-
Uruguay
-
Uzbekistan
-
Venezuela
-
Vietnam
-
Yemen
-
Zambia
-
Zimbabwe
- Forex
- Crypto
