On 8 January 2020, Facebook unveiled a new policy aimed at removing manipulated media that is designed to deceive viewers. The decision arrives amid escalating concerns over deepfakes—high‑quality video and audio fabrications that can make “seeing” unreliable.
The policy draws a line between two distinct forms of deceptive content. The first category targets media that has been altered in ways that an average user would not notice. This does not include obvious parody or clumsy edits, but rather subtle manipulations that make a person appear to say something they never actually said. Examples include audio that is synchronized to mismatched lip movements or a sentence constructed from fragments of different recordings.
The intent behind such edits is to mislead, and the tools required to produce them are becoming increasingly affordable and sophisticated. The second category focuses on material generated with artificial‑intelligence or machine‑learning techniques that replace, merge, or overlay elements onto existing video.
This covers face swaps, background alterations, or synthetic mouth movements that are matched to fabricated audio tracks. While similar effects once required Hollywood‑level resources, today a consumer‑grade laptop and a few hours of work can produce a fake that convinces a casual viewer. Monika Bickert, President of Global Policy Management at Facebook, presented these definitions as part of a broader enforcement framework.
The company stated that both deepfakes and other misleading manipulated media would be prohibited outright. The move is framed as a response to worries that such content could be weaponized—to tarnish political figures, spread false rumors, or trigger public panic—especially with the 2020 U.S. presidential election looming only months away.
Defining “deceptive” material presents a core challenge. Satirical edits are typically evident, whereas a propaganda clip that swaps a leader’s face onto another body may not be. Facebook’s policy therefore uses the “average person” test: if a typical viewer would be fooled, the content falls under the ban.
This subjective standard will inevitably be tested as the technology evolves. Enforcement is another hurdle.
With billions of users and a constant torrent of uploaded video, identifying every deepfake before it spreads is unrealistic. The policy functions as a statement of intent, promising action when flagged or discovered content violates the new rules. Yet the pace of AI‑driven creation tools outstrips that of detection technologies, meaning the policy may deter casual amateurs but could struggle against well‑funded, state‑backed disinformation operations.
In its announcement, Facebook acknowledged that the problem is not disappearing. AI tools that alter visual and audio content are advancing rapidly, and the platform’s ban applies to any material that appears authentic while conveying falsehoods.
Although this sets a high bar for enforcement, the company argues it is a necessary one. The policy rollout occurs amid broader pressure on social‑media platforms to curb misinformation. Critics have long accused Facebook of moving too slowly, and the deepfake ban represents a concrete, though limited, step in a continuing effort.
As manipulation technology improves, the platform will need to adapt its rules and detection capabilities accordingly. For now, Facebook has drawn a clear line: manipulated media that is likely to mislead will be removed.
Whether the company can consistently enforce that line remains to be seen.
