OpenAI removes accounts engaged in coordinated inauthentic behavior and influence operations
Policy change ● Confirmed 82% confidence first seen
OpenAI identified and banned multiple coordinated networks of accounts originating from China, Iran, and Ghana that were using its AI models to conduct surveillance planning, election manipulation, and influence operations across multiple platforms. The operations involved generating fake articles, social media content, and engagement metrics for purposes including election interference, political influence campaigns, and surveillance system development. OpenAI's enforcement actions and improved detection capabilities reflect the platform's efforts to combat coordinated inauthentic behavior and misuse of AI tools.
Decision brief
- What changed
- OpenAI disclosed four separate enforcement actions—naming operations 'Peer Review,' 'Sponsored Discontent,' and networks linked to Ghana's 2024 election and Iranian campaigns (IUVM, STORM-2035)—in which it banned accounts from China, Iran, and Ghana that used its models to draft surveillance proposals, generate fake articles/social posts, and fabricate engagement metrics for influence operations.
- Why it matters
- This confirms that generative AI tools are being actively exploited by state-linked and politically motivated actors for surveillance system design and cross-platform disinformation, not just hypothetically. For enterprises, this raises the stakes on AI platform trust, content provenance, and misuse detection, and signals heightened regulatory and reputational scrutiny around AI-generated content ahead of elections and geopolitical flashpoints.
- Evidence
- All four data points originate exclusively from OpenAI's own blog disclosures describing its internal threat-detection and enforcement actions; there is no independent third-party (e.g., government, academic, or rival platform) corroboration provided in this coverage, though the four cases are internally consistent in describing similar patterns of misuse and takedown.
- What remains uncertain
- The actual scale, reach, and real-world impact of these operations (e.g., how many people were exposed to the content, whether the Ghana election outcome was affected, or how advanced the China-linked surveillance proposals were) is not detailed. It's also unclear how these findings compare to misuse rates on competing AI platforms or whether OpenAI's detection captures the majority of such activity.
- Monitor next
- Watch for OpenAI's next periodic threat intelligence report or any independent verification from platforms, governments, or researchers confirming the scope and impact of these specific operations.
Analytical support, not advice — assumptions and open questions stated above.