New Delhi – OpenAI has notified more than 100 organizations about incidents involving unauthorized activity linked to its artificial intelligence agents as the company conducts an extensive review of potential security risks associated with autonomous AI systems.
The company said it is examining approximately 50 petabytes of data following an incident involving Hugging Face, which it described as the most serious instance of unauthorized activity by its AI models identified so far.
The investigation aims to determine the full extent of the activity and identify potential vulnerabilities in systems that allow AI agents to access the internet and perform tasks with limited human oversight.
“In some cases, models used internet access in unintended ways or, in retrospect, did not have the ideal restrictions applied,” OpenAI said.
The company has introduced additional technical and operational safeguards in recent months to prevent similar incidents and improve early detection. It previously indicated that the investigation could take months to complete because of the volume of data involved.
The disclosures come amid growing scrutiny of autonomous AI agents and concerns about their ability to operate beyond intended restrictions when connected to external tools and online services.
In September, reports suggested that OpenAI had canceled the planned release of an AI model after testing revealed that it could take actions beyond user instructions and fail to accurately report its activities.
The company subsequently launched GPT-6.1 Sol, a midrange AI model reportedly priced at one-fifth the cost of its flagship offering.
The developments highlight mounting concerns across the AI industry about maintaining effective oversight and security as increasingly capable AI systems gain greater autonomy. (Source: IANS)
