{ "id": 9, "name": "September 2026", "published": 1, "publish_date": "2026-09-24T10:00:00.000000Z", "pivots_populated_at": "2026-09-02T15:45:22.000000Z", "created_at": "2026-09-02T15:45:08.000000Z", "updated_at": "2026-09-24T10:00:03.000000Z", "type": 1, "chapters": { "9": { "id": 90, "name": "Crisis and Elections Response", "report_id": 9, "created_at": "2026-09-02T15:45:11.000000Z", "updated_at": "2026-09-02T15:45:11.000000Z", "commitments": [ { "id": 434, "name": "Elections 2025", "code": "50", "description": "[Note: Signatories are requested to provide information relevant to their particular response to the threats and challenges they observed on their service(s). They ensure that the information below provides an accurate and complete report of their relevant actions. As operational responses to crisis\/election situations can vary from service to service, an absence of information should not be considered a priori a shortfall in the way a particular service has responded. Impact metrics are accurate to the best of signatories’ abilities to measure them].", "report_id": 9, "field_type": "0", "order": 50, "created_at": "2026-09-02T15:45:11.000000Z", "updated_at": "2026-09-02T15:45:11.000000Z", "chapter_id": 90, "sig_value": "", "type_id": 6260, "specialMeasures": [ { "id": 2130, "name": "Specific Measures 50.1", "code": "sm50.1", "description": "Threats observed or anticipated", "report_id": 9, "commitment_id": 434, "field_type": "bool", "order": 1, "created_at": "2026-09-02T15:45:11.000000Z", "updated_at": "2026-09-02T15:45:11.000000Z", "sig_value": "
1. AI-generated content in electoral contexts: policy and enforcement gaps<\/strong> <\/div>
During the reporting period, DRI assessed how VLOPs govern AI-generated content in political contexts, as generative AI reshapes political communication and lets political actors create and spread synthetic content at scale and speed. VLOPs increasingly distribute this content, yet their policies remain inconsistent, opaque, and weakly enforced. We examined Meta, TikTok, and YouTube across five dimensions: actors, behaviours, content, degree, and effect. None of the three platforms mandate official political accounts to label AI-generated content, even though parties, candidates, and government actors reach larger audiences and carry greater authority than ordinary users. Most platforms do not treat AI-generated manipulative political content as a distinct electoral risk. TikTok addresses it under its integrity guidelines, but Meta and YouTube fold it into broader manipulated-media and misinformation frameworks. Automatic detection and labelling are applied inconsistently, with no clear triggers, leaving significant gaps in coverage. 

<\/div>
We tested these gaps empirically during Hungary's 2026 elections. Using TikTok's search function and election-related hashtags, we identified 2,043 posts from 22 accounts that reached over 50 million views and either posted predominantly AI-generated content or presented as AI-generated personas, focused on Hungarian politics. Nearly three quarters of the posts (71.6 per cent) contained AI-generated content, and 37.5 per cent of that content was left unlabelled by the user or the platform. We reported 199 unlabelled posts to TikTok for non-compliance with its Edited Media and AI-Generated Content labelling requirements. The company removed 19 videos and one account and restricted the visibility of two posts. The remaining 177 posts stayed unlabelled despite our flagging.  

<\/div>
Additionally, we submitted comments to Meta's Oversight Board on generative AI use during the elections. We argued that Meta's labelling standard sets a higher bar than the DSA Electoral Integrity Guidelines and called on Meta to align with the guidelines, cover the full electoral cycle, and sanction non-disclosure. 

<\/div>