Recycled video still beat AI in war falsehoods, and five-year-old Pegasus numbers came back as news

BOOM, the Indian fact-checking organisation, published a report by Archis Chowdhury on 5 November covering January to July 2026 and drawn from a stated sample of 527 fact-checks. AI-generated or AI-edited material featured in three of every ten checks, up from two in ten the previous year. Within the 81 war-related fact-checks tied to the US–Israel–Iran fighting, AI featured in 38, or 46.9 percent — while old, recycled footage featured in 42, or 51.9 percent. Only eight items were both. The AI mix inside the war subset was 17 synthetic videos, eight deepfakes, five voice clones and four images. By platform, Instagram carried the highest AI share at 37 of 103 checks (35.9 percent), ahead of X at 107 of 368 (29.1 percent) and Facebook at 70 of 263 (26.6 percent). The sample is India-centric; the war subset is the part that travels.

The recycling BOOM measures was visible in real time this week. A cluster of Indonesian-language articles published 6 to 8 November, including one at Indie Banyumas dated 8 November, presented the Pegasus Project as a current global investigation and repeated the familiar figures: a list of more than 50,000 phone numbers, from which journalists identified more than 1,000 people in 50 countries. Those are the July 2021 Forbidden Stories and Amnesty International Security Lab figures, five years old, with nothing new attached. The genuinely new element underneath is narrower and unverified: the IT specialist Teguh Aprianto, in remarks circulated through an X post on 6 November, argued that an alleged interception in a wiretapping case involving President Prabowo Subianto could have come via Pegasus rather than through the operator’s network, and said he believes Indonesia possesses the tool, citing a tender he says was opened and investigated.

Volume figures from the security industry carry the same hazard. The threat-intelligence vendor SOCRadar, whose Iran–Israel conflict dashboard was timestamped 8 November, counts more than 600 distinct attack claims across more than 100 Telegram channels in fifteen days — a tally of assertions made by self-declared actors, not of confirmed incidents, published by a company with a commercial interest in the number being large. Against that, Saudi Arabia’s state wire SPA announced on 6 November a national framework, “AI Principles in Media,” setting out eight principles and asking practitioners and content creators to disclose AI-generated or AI-assisted material through labels, watermarks or metadata. No issuing authority, enforcement mechanism, penalty or compliance deadline has been identified. It is a published document, not yet a regime.

Assessment: The policy apparatus is being built for synthetic media while the measured problem is still, narrowly, an old clip with a new caption. That matters for where money goes: watermarking and detection do nothing against a 2019 video of a different war, and disclosure rules bind the compliant. Treat any count of “attacks” sourced to Telegram tallies as a count of boasts. And set a tripwire — if the 50,000 / 1,000 / 50-countries figures surface anywhere in the coming weeks attributed to a November 2026 investigation, the item is recycled and the outlet carrying it has not checked.