US artificial intelligence (AI) firm OpenAI has said that it will train its most advanced models without access to the live internet, citing current limits in alignment research, monitoring systems and internet infrastructure security
NEW DELHI — US artificial intelligence (AI) firm OpenAI has said that it will train its most advanced models without access to the live internet, citing current limits in alignment research, monitoring systems and internet infrastructure security.
The company said that the policy, already implemented, will remain until safeguards improve enough to justify broader internet access for frontier models.
OpenAI said internet connectivity, though useful for teaching models about browsing and web‑based skills, also creates opportunities for a capable, misaligned agent to affect systems beyond its control.
The firm said in a blogpost that it paused training and evaluation of its most capable models after an internal research model briefly exploited a misconfiguration to gain limited internet access on September 2.
It resumed reinforcement learning training and internal inference of its frontier models on October 1 after implementing additional safeguards and further, went on with a campaign to probe its research infrastructure for vulnerabilities that could allow models to escape isolated computing environments.
“Red-teaming experiments discovered previously unknown software vulnerabilities, as well as misconfigurations in our research infrastructure,” the statement said.
The company also detailed its new approach to model containment, that it now mandates that the limits on what an agent can access must hold even when the agent tries to circumvent them.
"We define a strict security boundary around the resources the agent is authorized to use and design the surrounding infrastructure to enforce that boundary—even when agents make sustained, sophisticated attempts to breach it," it added.
Anthropic and OpenAI had recently disclosed a list of incidents in 2026 involving their AI models acting in unintended ways, including attempts to hack third-party websites, intensifying concerns about the security risks of cutting-edge AI.
Anthropic has said that its Claude AI model carried out a series of unintended actions on external organisations, including some US government websites, prompting a White House directive for AI firms to secure their systems.