Reported networks
OpenAI threat reports explained
Short answer
- Networks disrupted and reported, February 2024 to October 2025
- Over 40 4
- Latest report confirmed
- 25 February 2026 6
- Cases in the June 2025 report
- 10, four with a likely Chinese origin 3
- Covert operations in the May 2024 report
- 5 1
- Highest Breakout rating in the May 2024 report
- Category 2 1
Reports confirmed for this page
| Date | Report | What it covered |
|---|---|---|
| 30 May 2024 1 | Disrupting deceptive uses of AI by covert influence operations 1 | Five covert influence operations disrupted in the previous three months 1 |
| 5 June 2025 2 | Disrupting malicious uses of AI: June 2025 2 | 10 case studies, including covert influence operations, cyber operations, a scam and an employment scheme 3 |
| 7 October 2025 4 | Disrupting malicious uses of AI: October 2025 4 | Cyber operations, scam networks, authoritarian-linked misuse and two covert influence operations 5 |
| 25 February 2026 6 | Disrupting malicious uses of AI 6 | Case studies, including a Chinese influence operator 6 |
How a case study is written
Each case gives the operation a name, such as Sneer Review, High Five or Helgoland Bite in the June 2025 report 3. It states origin with qualifiers. The June 2025 report used phrases such as likely China-origin activity, Philippines-origin accounts and a threat actor apparently originating from Russia 3. It then lists what operators asked the models to do and the platforms where content appeared, such as TikTok, X, Reddit, Facebook and Telegram 3.
Each case ends with the action taken. In the October 2025 report OpenAI listed account bans and, in several cases, indicators shared with industry partners 5. Its usage policies prohibit foreign or domestic election interference and say breaking the rules may mean a user loses access 8.
The Breakout Scale
OpenAI rates influence cases on the Breakout Scale, published by the Brookings Institution in September 2020 7. Category One operations spread only within one community on one platform. Category Three operations spread across multiple platforms and reach multiple communities. An operation reaches Category Six if it triggers a policy response or includes a call for violence 7.
In May 2024 OpenAI said none of the five operations it reported scored higher than Category 2 1. In June 2025 it rated Sneer Review at the low end of Category 3, subject to the engagement proving authentic, and STORM-2035 at Category 1 3. In October 2025 it rated the Stop News operation at Category 2, down from an earlier rating of Category 3 5.
Main findings across the reports
- Reach: OpenAI said in May 2024 that the five operations did not appear to have meaningfully increased their audience engagement or reach through its services 1.
- Low views: the June 2025 report said typical STORM-2035 tweets recorded 150 to 350 views and zero likes, shares or comments 3.
- No new capability: the October 2025 report found no evidence of new tactics or that its models gave threat actors novel offensive capabilities 5.
- Many tools: the February 2026 report said threat actors may use different AI models at different points in their workflow 6.
- Mistakes: some operators posted model refusal messages in public, according to the May 2024 report 1.
Links with other reporting bodies
OpenAI's cases overlap with platform reports. Meta's first quarter 2025 report linked an Iran network to STORM-2035 activity reported by OpenAI and Microsoft in August 2024 9. OpenAI's October 2025 report said indicators in its casework overlapped with a Trellix report 5. The May 2024 report said OpenAI shared threat indicators with industry peers 1.
Common questions
How many networks has OpenAI disrupted?
OpenAI said on 7 October 2025 that it had disrupted and reported over 40 networks that violated its usage policies since February 2024 4.
Did AI make these influence operations more effective?
OpenAI said in October 2025 that it found no evidence its models gave threat actors novel offensive capabilities 5.
What is the Breakout Scale?
A six-category scale for the impact of influence operations, published by the Brookings Institution in September 2020 7.
Do OpenAI reports name companies?
Sometimes. The June 2025 report linked the High Five operation to a commercial marketing company in the Philippines 3.
Rules and charges change. Check the official source linked on this page before you pay or travel. Spotted an error? Email hello@influenceopsrecords.com.
Sources
- OpenAI: Disrupting deceptive uses of AI by covert influence operations (30 May 2024)
- OpenAI: Disrupting malicious uses of AI: June 2025 (5 June 2025)
- OpenAI: Disrupting malicious uses of AI: June 2025 (June 2025)
- OpenAI: Disrupting malicious uses of AI: October 2025 (7 October 2025)
- OpenAI: Disrupting malicious uses of AI: an update (October 2025)
- OpenAI: Disrupting malicious uses of AI (25 February 2026)
- Brookings Institution: The Breakout Scale: Measuring the impact of influence operations (September 2020)
- OpenAI: Usage policies
- Meta: Adversarial Threat Report, First Quarter 2025 (May 2025)
Related
- AI-generated personas in influence operations
What Meta, OpenAI and the EEAS have reported about AI-made profile photos, fake newsreaders and AI-written text in influence networks, with dates.
- Meta's adversarial threat reports explained
What Meta's adversarial threat reports contain, how each network entry is laid out, and what the Q4 2024, Q1 2025 and Q2/Q3 2025 reports listed.
- Google TAG bulletins explained
How Google Threat Analysis Group bulletins list terminated YouTube channels, blogs, ads accounts and blocked domains, with 2025 examples.
- How attribution of influence operations works
How Meta, Google, OpenAI, Microsoft, X and the EEAS word attribution, from country of origin to named entities, and what terms like likely and linked mean.