Facebook's deepfakes policy is a good move but remains deeply flawed, covering only certain videos created by AI and creating loopholes for misleading “satire”
good old fakes—will easily slip by annoticed. And how many writers on propaganda have warned that irony/satire are anyone's game & a hair's breadth away from disinfo? https://www.theverge.com/... @aelkus : irrespective of this particular point its time to also maybe consider that the actual problem itself is intractable https://twitter.com/... Will Oremus / @willoremus : And yet! For all its limitations, Facebook's policy is miles better than what other social networks are doing—which is embracing deepfakes as a feature and building tools for users to make their own. https://onezero.medium.com/... https://twitter.com/... Will Oremus / @willoremus : Making an exception for deepfakes that are “parody or satire” seems like a defensible position. But it may also be highly exploitable. As I reported last month, much of the viral fake news on the platform today uses the “satire” tag as a fig leaf. https://onezero.medium.com/... https://twitter.com/... Will Oremus / @willoremus : Facebook's ban on deepfakes is even narrower than it seems. My close read: https://onezero.medium.com/... https://twitter.com/... Charlie Warzel / @cwarzel : New from me on how Facebook and Twitter and the dangerous “newsworthiness” loophole that lets public figures act recklessly. Platforms claim it's in the public interest. Maybe. But newsworthiness is a choice masquerading as an inevitability. https://www.nytimes.com/... Andy Stone / @andymstone : I shared inaccurate information earlier. I sincerely regret the error and need to correct the record. With regard to our new policy, whether posted by a politician or anyone else, we do NOT permit in ads manipulated media content, as defined here: https://about.fb.com/... Berkeley Engineering / @cal_engineer : Facebook's new deepfake policy: ban the high quality fake videos, keep the crappy ones. “Why focus only on deep-fakes and not the broader issue of intentionally misleading videos?” Berkeley prof Hany Farid said https://www.washingtonpost.com/ ... #deepfake David Uberti / @daviduberti : Biden campaign, which has been vocal on Facebook/disinfo, calls new deepfake ban an “illusion of progress. Facebook's policy does not get to the core issue of how their platform is being used to spread disinformation, but rather how professionally that disinformation is created.” https://twitter.com/... Hadas Gold / @hadas_gold : Update: Facebook just told me they made a mistake when telling me this earlier. Deepfakes in a policial ad will NOT be allowed per policy “Whether posted by a politician or anyone else, we do not permit manipulated media content in ads” - big difference ! See also Mediagazer
Context & Ripple Effects
Facebook's new rule arrives after months of internal debate: back in June, Mark Zuckerberg argued the definition of deepfakes was too broad and needed to be scoped down, which is exactly what this policy does — it covers only certain AI-generated videos while carving out an exemption for "satire" that critics say propaganda writers have long exploited. The narrower the definition, the more it excludes: the Washington Post notes the policy may not even reach simple distortions like the doctored Nancy Pelosi video that needed no AI at all.
The contrast with peers sharpens the critique. Twitter's draft policy from November chose labeling and share warnings over bans, while OneZero points out other networks are embracing synthetic media outright as a creation feature. Against that backdrop, Facebook's ban looks both ahead of rivals and structurally incomplete.
First-order effects
- Political operatives and disinformation actors now have a documented playbook: keep manipulations non-AI-based or frame them as satire, and Facebook's enforcement won't touch them — the Pelosi-style distortion remains postable under the new rules.
- Facebook's policy team, including spokespersons like Andy Stone, inherits the burden of defending why AI-made fakes are banned while cheaper, equally misleading edits are not.
Second-order effects
- Twitter faces pressure to harden its label-and-warn approach: if Facebook bans outright, Twitter's lighter-touch draft risks becoming the platform of choice for synthetic political media that Facebook rejects.
- The policy's narrowness pushes weight onto detection tooling — yet Facebook's own Deepfake Detection Challenge later showed its winning algorithm catching fakes at only about 65% accuracy, meaning enforcement gaps can't simply be closed by better classifiers.
Third-order effects
- If platforms each draw their own line between 'manipulated' and 'satirical' media, synthetic political content migrates to whichever network defines deception most narrowly — fragmenting moderation into a lowest-common-denominator problem rather than solving it.
- The pattern points toward regulation or shared standards stepping in where platform-by-platform definitions fail, since a rule scoped by production method (AI vs. manual edit) rather than effect (deception) leaves the actual harm unaddressed.
The trend: Platform governance of synthetic media is converging on narrow, enforceable definitions that trade completeness for clarity, pushing the hardest cases — cheap fakes and labeled satire — out of policy and into detection arms races.