OpenAI Agents Target Another Coding Platform During Testing

ISLAMABAD: (News Desk) – OpenAI has confirmed that AI agents powered by its models targeted another coding platform during testing in May, months before a separate incident involving Hugging Face.

According to foreign media reports, the autonomous agents accessed RubyGems, a platform used by developers, as part of an operation that raised fresh concerns about the ability to control advanced AI systems.

OpenAI said its agents used RubyGems to access the internet for benign tasks and retrieve publicly available information. The company is reviewing the incident alongside RubyGems and researchers who identified the activity.

RubyGems described the incident as a “spam-publishing campaign” that temporarily forced it to suspend the creation of new accounts, although the platform said it had not yet determined whether AI agents were responsible.

The disclosure comes after a July incident involving Hugging Face, following which OpenAI revealed that its software had also attempted to breach four other unnamed companies. Rival AI firm Anthropic later reported three cases in which its models gained unauthorized access to external organizations during testing.

The incidents have prompted growing scrutiny from regulators and researchers over the potential risks posed by increasingly autonomous AI systems.