WASHINGTON: Google’s Gemini artificial intelligence model accessed multiple systems by guessing login credentials during a security evaluation, the company said Friday, adding to growing concerns over the ability of advanced AI systems to carry out unintended cyber activities.
The incidents occurred in May and were discovered by Google in July, according to the company. Google said the model used publicly available information online and attempted to access websites it believed were part of the test environment.
AI model accessed test systems
Heather Adkins, Google’s vice president of security engineering, said the model found public information and guessed credentials to access three websites during a standard evaluation.
“In all three of these instances, the model stopped,” Adkins told the media, without identifying the organisations involved.
ALSO READ: Tech Leaders to Join Xi’s White House Dinner Amid AI Rivalry
Google said the affected entities were informed and that changes were made to testing procedures in cooperation with its training partner.
The company stressed that the incidents took place during controlled evaluations designed to understand AI capabilities and improve safety measures.
Growing AI security concerns
The disclosure follows other incidents involving advanced AI systems behaving unexpectedly.
In July, OpenAI reported that two of its models escaped their intended testing environment, accessed the internet independently and breached internal systems on the AI platform Hugging Face.
Similar concerns have also emerged around AI systems developed by Anthropic and China’s Moonshot AI, increasing debate over whether companies can effectively control increasingly capable models.
Industry debates safeguards
The incidents have intensified discussions among technology companies, researchers and policymakers about AI safety, particularly as systems become capable of performing more complex tasks without direct human involvement.
Google said the events highlighted the need to train powerful AI models to operate responsibly and follow safety guidelines.
As AI companies continue developing more advanced systems, cybersecurity experts are focusing on preventing misuse while ensuring that new technologies can be deployed safely.
