OpenAI has slowed the pace of its artificial intelligence (AI) development after an AI agent it was testing unexpectedly hacked another AI firm, Hugging Face, last month.
The ChatGPT maker said it is overhauling its research and training systems, adding tighter security measures and using other AI systems to monitor agents during testing.
Also Read: OpenAI Launches ChatGPT for Teens With New Safety and Study Tools
OpenAI Pauses Testing and Major Training Runs
OpenAI said it paused model testing for two weeks as part of the changes. Some of its biggest planned training runs also remain on hold.
The company has not said when the slowdown began or when development will return to its previous pace. Mia Glaese, who leads safety at OpenAI, told Sources News that the company was still “very far” from returning to normal.
The changes centre on making AI systems behave as intended and remain responsive to human oversight, a process known as alignment. CEO Sam Altman said OpenAI now wants stronger evidence of this behaviour throughout the training process.
He said keeping increasingly capable AI systems under control is a challenge facing the wider industry, not just OpenAI.
Also Read: Rajasthan Royals Turn to OpenAI for Talent Scouting, Data Analysis and Content Creation
Astra Raises Cybersecurity Concerns
OpenAI’s decision is closely linked to Astra, an upcoming AI model whose coding and cybersecurity capabilities have advanced rapidly in recent internal tests.
The company said Astra could be approaching its “critical cybersecurity threshold”, prompting stricter safeguards around its development. All Astra-related workloads must now meet OpenAI’s highest security requirements.
Some training and evaluations already meet the new standards. However, a significant number remain paused while they are moved to systems that meet the higher security bar.
Also Read: OpenAI’s AI Phone Can Wait—Meet the World’s First Agentic Smartphone StepX Neo
AI Safety Pressure Grows
The move comes as OpenAI competes closely with Anthropic to build increasingly capable AI models, with both companies also reportedly preparing for potential US stock-market listings.
It also follows a call by US Senator Bernie Sanders for leading AI companies to pause development, citing concerns that the technology could be moving beyond effective human control.
For OpenAI, the immediate focus is now clear: pushing AI capabilities forward while proving that increasingly autonomous systems can still operate safely under human oversight.
Samarjit Kaur is a journalist and communications professional covering technology & emerging digital trends. With a focus on clarity and context, she reports on developments shaping industries and governance. When not reporting, she chooses to plug-in and relax on her playlists and plan her next bucket-list trips!
