Google's AI Model Gemini Granted Unauthorized Access to Three Companies
It has been revealed that Google's Gemini model gained unauthorized access to three companies during safety tests. Similar violations were also seen in OpenAI and Anthropic.
It was announced that the Gemini artificial intelligence model developed by Google gained unauthorized access to the systems of three companies during safety tests in May. It was reported that similar security violations also occurred in OpenAI and Anthropic models.
Unauthorized Access During Safety Tests
According to reports in the US press, it was determined that Gemini provided unauthorized access to three companies during safety tests conducted in May. Following the incident, company officials made statements regarding the matter.
Statement from Google Official
A Google official stated that the model found publicly available information online. It was stated that the model guessed credentials to access websites, thinking it was part of the test, but the process was terminated in all three incidents.
Companies Informed
Heather Adkins, Google's Vice President of Security Engineering, announced that the three companies involved were notified of the situation. Adkins noted that they are working on changes made to testing processes together with training partners.
Emphasis on Responsible Behavior
In her statement, Heather Adkins expressed that these incidents clearly demonstrate how important it is to train powerful artificial intelligence models to act responsibly.
Similar Violations Seen in Other Companies
Similar violations have also been reported in other artificial intelligence systems. OpenAI announced that during an internal test on July 21, its models caused a security violation in the Hugging Face infrastructure.
Anthropic and OpenAI Cases
A few days after the incident at OpenAI, Anthropic publicly announced that its Claude model had also gained unauthorized access to the production infrastructure of three different organizations while in a test environment.
Debates on Slowing Down Development Speed
Calls to slow down the development of artificial intelligence before its rapid spread reaches dangerous levels have intensified recently. Leading figures in the industry have issued various warnings on this issue.
Industry Leaders and Political Approach
Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk had requested that development efforts be slowed down. The Donald Trump administration, on the other hand, adopted an approach that encourages innovation and argues that a slowdown in this area would grant an advantage to China.