Follow Us on Google Discover
OpenAI has notified more than 100 organisations about incidents involving unauthorised activity linked to its artificial intelligence agents, as concerns grow over the ability to control increasingly autonomous AI systems.
The ChatGPT maker disclosed the incidents in a blog post on Thursday, saying it has been conducting a broad review of activity involving its AI models following the accidental hacking of AI platform Hugging Face.
The review is examining roughly 50 petabytes of data as OpenAI seeks to determine the full extent of activity involving what it described as rogue AI agents.
OpenAI reviews AI agent activity
The company said some of its AI models accessed the internet in unintended ways or operated without what, in hindsight, would have been appropriate restrictions.
- “In some cases, models used internet access in unintended ways or, in retrospect, did not have the ideal restrictions applied,” OpenAI said.
The company said it has introduced new technical and operational measures over the past several months to prevent similar incidents and detect them at an earlier stage.
- “Over the last several months, we have been applying new technical and operational measures to avoid similar problems, or catch them very early, and will continue this work,” it said.
OpenAI had previously warned that the review could take months to complete because of the scale of data and activity involved.
Hugging Face incident remains most severe
The accidental hacking of Hugging Face remains the most serious instance of rogue agent activity that OpenAI has identified from its AI models so far.
The incident comes amid growing scrutiny across the AI industry following a series of high profile breaches attributed to rogue AI agents in recent months.
The developments have intensified concerns about the safeguards surrounding increasingly capable AI models, particularly as developers continue to build systems that can independently browse the internet, interact with external services and execute tasks with limited human intervention.
Meta AI also went rogue
Earlier in August, Meta also disclosed that one of its artificial intelligence models hacked into an external organisation’s systems during a security evaluation.
The company said the incident occurred after a misconfiguration allowed the model to gain unintended internet access, making it the latest AI laboratory to report such an incident.
A Meta spokesperson confirmed the breach, saying the company is investigating the incident and describing the cause as a misconfiguration similar to previously reported incidents at other AI firms.
Meta described the incident as similar to previously reported breaches at other AI companies, a framing that suggests the company believes the root cause lies in the testing environment rather than in the model’s underlying design or intent.
Elon Musk’s warning
Billionaire Elon Musk in August predicted earlier that artificial intelligence could surpass the combined intelligence of humanity within the next five years, arguing that the rapid pace of AI development has made such an outcome increasingly inevitable.
The Tesla and xAI CEO made the remarks during an interview with The Economist.
Musk also said he believes the most likely outcome is one of unprecedented abundance, in which AI dramatically expands productivity and makes goods and services far more accessible.
