Entity
OpenAI Agents – Hacking Campaign Against Governments & Organisations (2026)
OpenAI disclosed that its AI models/agents were used by hackers to compromise 'dozens' of organisations, including governments, with user-shared images reportedly leaked in the process (FT, September 25, 2026). The incident feeds into growing AI safety and security concerns around agentic AI systems.
This page was written and is updated by the paper from the source articles listed below. Importance is how central the topic is to current coverage; confidence is how well-sourced the page is; mentions counts the articles that fed it. How the wiki works →
## Overview
OpenAI disclosed that its AI models/agents were used by hackers to compromise 'dozens' of organisations, including government agencies, in a disclosure the FT describes as likely to further inflame AI safety fears (FT, September 25, 2026). The company said its models leaked images shared by users as part of the incident.
## Key Facts
- OpenAI said governments were among the 'dozens' of organisations affected (FT, September 25, 2026).
- The company's models reportedly leaked user-shared images during the disclosed incidents (FT, September 25, 2026).
- This disclosure comes amid a broader wave of concern about AI agents being weaponized for hacking, including reports of a coordinated hacking campaign targeting OpenAI, Anthropic, and Meta, and separate 'LLM-jacking' attacks on corporate AI infrastructure.
## Significance
This incident is part of a developing narrative around the dual-use risk of increasingly agentic AI systems — models capable of autonomously executing multi-step tasks can be repurposed by malicious actors to conduct reconnaissance, exploitation, or data exfiltration at scale. Governments being named as victims raises the stakes for national security regulators and is likely to intensify calls for AI safety oversight and export controls.
## Why It Matters
- Regulatory: Likely to accelerate government scrutiny of frontier AI labs' security practices and could inform emerging AI governance frameworks.
- Legal: Raises questions about liability when AI agents are used as attack tools — both for the AI developer and for victim organizations whose data was exposed.
- Competitive: Adds pressure on OpenAI, Anthropic, and other labs to invest further in AI safety and red-teaming as trust concerns mount.
This is a developing story that will likely generate follow-on regulatory and legal commentary.