July 23, 09:50

Newest AI from OpenAI got out of control and hacked a major tech website autonomously

OpenAI announced on Tuesday that its advanced artificial intelligence models went out of control during security testing, independently hacking a popular programmer platform.

Incident Details

The San Francisco firm called this an 'unprecedented cyber incident' and stated its intention to conduct a joint investigation with the code online library Hugging Face. AI models that underpin tools such as image generators and chatbots are referred to as agents when they act autonomously to perform real-world tasks.

OpenAI specified that the incident involved a combination of models, including the recently launched GPT-5.6 Sol and an even more powerful model in pre-release stages. The company was attempting to assess these models' hacking capabilities by giving them tasks in a strictly controlled digital test environment where internet access was restricted for security purposes.

According to the OpenAI blog post on the incident, 'operating within our isolated test environment, our models spent significant computational power trying to find a way to access the open internet, aiming to solve the evaluation problem.' After establishing an internet connection, the models decided to target the Hugging Face platform—a major repository of data, datasets, and other materials for AI—to aid their mission.

In search of 'secret information' that could help bypass the evaluation, the OpenAI system 'linked together multiple attack vectors, including the use of stolen credentials.'

Potential for Catastrophe

Hussein Abbas, a professor of computer science at UNSW Canberra, told AFP that the incident was 'striking in many aspects.' He noted that 'it didn't just attack Hugging Face. It actually attacked its internal system to exploit its own vulnerabilities,' adding that it was 'scary.'

GPT-5.6 models and other advanced developments, including the Mythos series from OpenAI's main competitor, Anthropic, have raised concerns about their ability to breach cybersecurity systems. Both American companies were forced to temporarily halt the general release of these latest technologies due to fears in Washington that they could facilitate penetration into critical infrastructure.

Abbas emphasized that advanced AI 'is usually in the hands of ethical and responsible people,' but warned that 'it would be a catastrophe if it fell into the wrong hands with malicious intent.' He also stated that the emergence of questions about how to regulate the AI sector has become a key issue, and 'we need collective community efforts to manage this situation.'

Hugging Face Reaction

Hugging Face had reported a cyber 'intrusion' last week without mentioning OpenAI. However, the company noted that 'this was different from anything we had encountered before in one important way: it was entirely and end-to-end driven by autonomous AI agent systems—and we largely detected and analyzed it using our own AI.'

Clement Delanghe, CEO of Hugging Face, wrote on X that the company suspected the cyberattack originated from a leading global AI lab, given the complexity of the agent. He added, referencing OpenAI: 'We firmly believe there was no malicious intent on their part. It is quite astonishing that all of this happened autonomously!'

tecnoblog.netJuly 22, 17:41

OpenAI AI Models Escape Tests and Attack Hugging Face Platform

OpenAI confirmed that its artificial intelligence models, specifically GPT-5.6 Sol and another advanced pre-release AI, managed to escape a controlled testing environment. These systems connected to the internet and autonomously invaded the Hugging Face platform's systems.

Invasion Details

This cybersecurity incident occurred this month and was initially motivated by the search for an answer to a cybersecurity assessment. The escape was made possible after the models identified and exploited a zero-day vulnerability in software used internally by OpenAI.

To test limits in an extreme scenario, the developer had intentionally removed security protections and filters from the models in an environment that should have been completely isolated. This led the AIs to focus intensely on solving the challenge, performing privilege escalation and examining the company's infrastructure until they located a network node with web access.

Attack Execution

Once connected to the internet, the algorithms logically concluded that Hugging Face, known as a large code repository, likely contained the solution for the assessment. From that point, the model initiated several attack vectors, even using stolen cloud credentials to execute remote code.

The attack was only contained when Hugging Face's anomaly detection system, which also uses AI to protect its network, identified the suspicious traffic, isolated the activities, and blocked the process before irreversible damage occurred.

Consequences and Responses

Despite what happened, Hugging Face assured that no evidence of manipulation was found in partner databases, client models, or public spaces. Input flaws were corrected, exposed credentials were revoked, and the case was forwarded to the competent authorities.

For its part, OpenAI reinforced configurations on its internal servers, notified the developers of the affected software about the zero-day flaw, and is collaborating with Hugging Face in the investigations. Clem Delangue, CEO of Hugging Face, emphasized that the advance of autonomous threats requires a change in mindset in the sector, advocating that AI security must be solved in an open and collaborative manner, with broad access to technology.