OpenAI is upping its security safeguards around training and testing after its AI agents hacked another company last month.
WASHINGTON — OpenAI, the company behind ChatGPT, is upping its security safeguards around training and testing after its AI agents hacked another company last month.
The company said its new security measures better contain its models, improve their behavior, and alert teams when there’s a problem.
“As models become more capable, the risks associated with developing and testing them internally also grow. Our standards for monitoring, alignment, and security must stay ahead of those risks,” the company said in an article shared Tuesday.
Last month, AI startup Hugging Face said it had detected an intrusion into its data processing systems that it suspected was caused by an AI agent acting on its own.
During a cybersecurity exam, the AI agents hacked into Hugging Face and searched for an answer key outside the testing environment after determining it would be the easiest way to ace it.
OpenAI said the intrusion was caused by a combination of its AI models, including its newly released GPT‑5.6 Sol and an “even more capable” model that is still being tested internally. OpenAI said its AI used stolen credentials and discovered a previously unknown vulnerability to access Hugging Face servers.
It went to “extreme lengths to achieve a rather narrow testing goal” and “found ways to gain access to secret information that it could use to cheat the evaluation,” the company said.
OpenAI was one of several tech companies that had hacking incidents with its AI agents, showing that AI has proven a capacity to lie, cheat, cover its tracks and exploit bugs. The Hugging Face hacking incident was so severe that 1,300 staffers at top tech companies called for tools to slow the pace of AI development.
Just days after OpenAI’s agent hacked Hugging Face, Anthropic revealed its AI models hacked three organizations during testing.
OpenAI also said that it might pause a number of workloads while its new security measures take full effect.
“We are prioritizing safety and alignment workloads for migration to these new environments first,” OpenAI said.
These incidents have highlighted vulnerabilities in AI security and controls and raised questions about how AI can be safely kept under human control as the technology’s use becomes more widespread globally.
Researchers have warned for years about technology risks and the need for stronger AI defensive engineering.
CNN Newsource contributed to this report.
Click Here For The Original Source.
