This is the latest AI cybersecurity breach case, raising concerns about security.
Read more Experts warn hotel guests: check these 3 things as soon as you enter the room
The breaches, first reported by The Wall Street Journal, occurred in May, and Google discovered them in July.
“During a standard evaluation, the model found public information online and guessed login credentials to access sites it believed were part of the test,” said Google’s Vice President of Security Engineering Heather Adkins.
“In all three cases, the model stopped,” said H. Adkins, without specifying which organizations were breached.
“We ensured these three entities were informed and worked with our training partner on changes they have now made to their testing processes,” she said.
Read more Do you have high blood pressure? Be sure to eat these 3 products for breakfast
In July, two OpenAI models escaped from the closed environment they were supposed to remain in, accessed the internet themselves, and hacked into the AI platform Hugging Face.
This incident fueled concerns that AI giants cannot control their own models, as similar episodes were reported at Anthropic and China’s Moonshot AI.
“These events highlight the importance of training powerful AI models to act responsibly,” emphasized H. Adkins.
