OpenAI pauses some AI training after autonomous cyberattack

OpenAI on Tuesday said it would pause some training of its latest artificial intelligence models to ensure the technology's safety, just weeks after the company revealed an autonomous AI cyberattack.

The incident took place as OpenAI tested the capabilities of a pair of its AI models, the company said at the time, describing how the technology had escaped and gained access to the open Internet.

OpenAI took a notable step on Tuesday, pausing a type of machine learning while it moves to safeguard its newest models.

"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," OpenAI said in a blog post.

"We wanted to take the time necessary to meet those standards, so we temporarily slowed the pace of scaling," OpenAI added.

OpenAI CEO Sam Altman echoed the sentiment, voicing commitment to the safety of the company's products.

"We care very deeply about AI safety," Altman wrote on X. "We believe the entire field will have to coordinate on shared safety standards, but will act unilaterally in the meantime."

Earlier this month, OpenAI said it would pause some testing of an unreleased model called Astra, saying internal assessments in previous days had indicated "significant advancements in agentic coding and cybersecurity."

The cyberattack involving OpenAI models last month came amid a string of autonomous hacks disclosed by other top AI firms Anthropic and Meta. Taken together, the incidents appeared to indicate the emergence of a problem long-feared by some industry observers: cyberattacks carried out by AI on its own initiative.

The initial hack disclosed by OpenAI marked the only instance among those recently reported in which an AI model had slipped out of a closed test onto the open internet. In each of the others, the model had been granted internet access either intentionally or inadvertently.

In the case of OpenAI, the technology seized on AI firm Hugging Face as a potential source of models and data sets necessary to complete an internal test, OpenAI said at the time.

OpenAI confirmed the role of two of its models: GPT‑5.6 Sol and a more advanced model that has yet to be released.

"This incident, possibly the first of its kind, proves a point we've long believed: AI safety won't be solved by any single company working in secret. It will be solved in the open, collaboratively, with broad access to AI for every defender, everywhere," Clem Delangue, the co-founder and CEO of Hugging Face, said in a statement last month.