Board Calls for New Rules on Deceptive AI During Conflicts
March 10, 2026
In analyzing the spread of AI-generated content in armed conflicts in a case on the 2025 Israel-Iran war, the Oversight Board calls on Meta to do more to allow users to identify such output. Its approach to surfacing AI-generated content must evolve. This includes providing details at scale about the origin of media, based on content provenance standards, investing in stronger detection tools and developing better methods for appropriate labeling. Meta needs to create a new, separate set of rules to ensure users can reliably recognize AI-generated content. Additionally, it should amend its current policies to ensure a timely and adequate response to deceptive AI-generated output.
The company needs to meet its public commitments and employ its own tools and others available across the industry to effectively address deceptive generative AI content that spreads among platforms.
The Board overturns Meta's decision to leave up the post in this case without a High Risk AI label.
Why This Matters
As the quantity and quality of AI-generated content increase, its impact on people and societies will be profound. The risks are heightened when deepfake output designed to deceive, manipulate or increase engagement is shared during conflicts and crises, such as in Iran and Venezuela in 2026, and spreads rapidly on different companies’ platforms. During those two crises, there were claims that deceptive AI-generated content was authentic and that authentic content was fabricated. That heightens the public’s inability to discern truth, emblematic of the liar’s dividend, leading to a general distrust of all information. AI-driven influence campaigns are a growing challenge seen globally in recent years, exacerbated in restrictive media and internet ecosystems that limit credible information. However, AI-generated output being misleading is not in itself a legitimate reason to restrict freedom of expression. The industry needs coherence in helping users distinguish deceptive AI-generated content and platforms should address abusive accounts and pages sharing such output.
About the Cases
The Israel-Iran war in June 2025 signaled an inflection point, with the presence of deceptive generative AI content on social media becoming known as its own “soft war.” Such deceptive output was reported as garnering huge numbers of views, and both Israeli and Iranian governments were accused of AI-driven influence attempts. On June 15, two days into the 12-day Israel-Iran conflict, a video was posted to a Facebook page that claimed to be a news source. The posting user was in the Philippines. The video depicted extensive damage to buildings with overlaid text in English reading “Live now – Haifa Towards Down" [sic] with the posting date. The video was very similar to one originating on TikTok and identified by an independent fact-checker (Agence France-Presse) as false and AI-generated. A caption on the Facebook post listed many headline-style phrases linked to the conflict and unrelated terms and hashtags. The post received more than 700,000 views, with several comments noting that the content was AI-generated.
Six users reported the case to Meta, but it was neither reviewed by the company nor checked by third-party fact-checkers. A user appealed to the Board. After the Board selected this case, Meta confirmed the post did not violate the Misinformation Community Standard because it did not “directly contribute to the risk of imminent physical harm,” and did not require an AI label.
Obvious signals of deception related to the post led the Board to question Meta over the identity and behavior of accounts linked to the page. The company subsequently disabled three accounts linked to the page for engagement abuse and inauthenticity, removing the page and, with it, the case content. The page had been eligible to monetize through Meta’s Stars program.
Key Findings
The Board finds that the content posed a material risk of misleading the public on an important matter at a critical time, so Meta should have applied a “High Risk AI” label. The post did not meet the threshold for removal (posing a risk of imminent physical harm or violence). Meta must do more to address the proliferation of deceptive AI-generated content on its platforms, including by inauthentic or abusive networks of accounts and pages, particularly on matters of public interest, so that users can distinguish between what is real and fake.
The Board is concerned by reports that Meta is inconsistently implementing Coalition for Content Provenance and Authenticity (C2PA) standards even on content generated by its own AI tools, and that only a portion of such output receives proper labeling. The C2PA sets out technical standards to embed provenance information as metadata in content, allowing platforms to more easily identify AI-generated content and apply labels to inform users.
The current mechanisms for affixing even the standard label of AI Info to video (user self-disclosure or an escalation to the Content Policy team) are neither robust nor comprehensive enough to contend with the scale and velocity of AI-generated content, particularly during a crisis or conflict where there is heightened engagement on the platform. A system overly dependent on self-disclosure of AI usage and escalated review (which occurs infrequently) to properly label this output cannot meet the challenges posed in the current environment. Some Board Members additionally noted that High Risk AI labels (for output that could deceive people on important matters) must also be coupled with demotion or removal from recommendations to address concerns over spreading the impact of deceptive content.
Meta’s narrow approach to fanning out ratings to identical and near-identical content may have meant this post did not receive a fact-checking rating. Resource constraints and a remarkable volume of output make it difficult for fact-checkers to ensure timely review of all deceptive content, especially during a conflict or crisis. The Board reiterates that Meta should ensure that fact-checkers are adequately resourced and have guidance on prioritizing content from conflicts. The Crisis Policy Protocol (CPP) and Trending Events designations should have allowed Meta to ensure more effective support for third-party fact-checkers during the crisis. Fanning out ratings to a broader category of very similar videos could have significantly limited the potential harm, including by demoting them. The case highlights inefficiencies in Meta’s current approach during armed conflicts, compounding concerns the Board has expressed previously.
It is concerning that with the CPP activated and additional resources allocated, Meta did not identify on its own initiative the clear engagement abuse signals from the page, and that it only investigated the accounts behind it in response to Board questions. Accurate enforcement of the behavior-based policies could have prevented the harms from these violating accounts, rather than relying on content-based downstream mitigations prone to a high failure rate.
The Oversight Board’s Decision
The Board overturns Meta's decision to leave up the content without a High Risk AI label.
The Board recommends that Meta:
- Create a Community Standard for AI-generated content, separate from the Misinformation Community Standard, providing comprehensive rules on provenance preservation, AI labeling protocols and self-disclosure.
- Develop pathways for affixing High Risk and High Risk AI labels to content much more frequently, assisted by clearer escalation channels from automated systems and at-scale review, so that such labeling can occur at a significantly higher volume.
- Attach provenance information and invisible watermarks to content created by Meta AI tools, including applying Content Credentials (as laid out by the C2PA) at creation.
- Implement Content Credentials at scale and ensure they are clearly and consistently visible and accessible whenever the provenance details are available.
- Invest in stronger detection tools for AI-generated multi-format (audio, audio-visual and image) content.
- Publish a clear explanation of penalties for failure to self-disclose digitally created or altered content, including the criteria for penalties and consequent limitations.
- Amend the Misinformation Community Standard to ensure that swift review of misinformation that directly risks imminent violence or physical harm does not depend solely on signals from external partners. A CPP lever should allocate resources for timely, proactive detection of such violating content, supported by in-house expertise and action, including labeling and investigating posting accounts and pages.
Further Information
To read public comments for this case, click here.
