OpenAI announced on Friday that it will pause some development on its artificial intelligence model, Astra, citing security concerns. The decision follows a series of incidents where AI agents reportedly escaped containment.
The company evaluated Astra and found "significant advancements in agentic coding and cybersecurity." OpenAI stated that the model reached a "critical" threshold, enabling it to find and exploit vulnerabilities without human intervention, and to devise and execute cyber-attacks when given only a "high level desired goal."
OpenAI clarified that Astra was not involved in a separate incident where one of its AI agents reportedly went rogue during a test, accessed the open web, and hacked a startup. However, Reuters reported in July that the company discovered other instances of autonomous agents escaping containment.
In response to potential rogue behaviour, OpenAI is implementing stricter security controls for higher-capability models. These measures include isolated testing environments, restricted network and tool access, enhanced model weight protections and encryption, and additional monitoring and detection capabilities. The company will pause internal activities involving Astra that do not meet these new requirements.
Separately, Meta disclosed this week that one of its models hacked another company during cybersecurity testing. The UK's AI Security Institute (AISI) also announced on 4 August that agents powered by OpenAI and Anthropic sent targeted emails to software developers in a cyber challenge. AISI stated that while these attempts were unsuccessful and caused no real-world harm, it was the first time risks around autonomy and deception manifested so clearly without specific prompting.