Key Takeaways:
Recycled real footage presented as evidence of 2026 midterm machine manipulation, plus OpenAI’s 8 October disruption of two AI-assisted influence operations using false identities, show how genuine material can carry false claims.
The piece argues that public warnings must separate verified facts from attribution, disclose evidentiary limits and use a two-stage process of preliminary notices followed by stable, revisable evidence records.
Effective safeguards include independent oversight of secret material, visible correction histories that travel as far as original accusations, and protection of free expression.
AI-enabled false fronts and recycled election footage show why public warnings must distinguish verified facts from attribution, disclose their limits and remain open to correction.
In the days around the 2026 U.S. midterm elections, videos appeared online purporting to show voting machines switching voters’ selections. The footage was not invented by an AI image generator. It was real footage from earlier years, presented as evidence of a new election problem. A November 2 Reuters verification identified the old devices and dates and explained why the clips did not establish manipulation in the current election.
The lesson travels well beyond the United States. A genuine recording can carry a false accusation when its time, location or context is replaced. Conversely, a carefully documented investigation can become misleading if its findings are stretched into claims about perpetrators or motives that the evidence cannot establish.
On October 8, OpenAI reported disrupting two AI-assisted influence operations that used false institutional identities. One Russia-origin network sought credibility through a purported Latin American research organization and apparently unwitting local participants. An Iran-origin operation used fabricated journalist personas to pitch articles to outside outlets. The company described activity visible through its services and related investigation; that visibility should not be confused with independent proof of every claim made by the operators, every audience effect or the ultimate chain of state direction.
These episodes raise a question for governments that warn the public about influence campaigns: how much of a warning can citizens check, and what happens if official attribution later needs correction? Democracies must expose genuine manipulation without making government authority a substitute for evidence.
In the days around the 2026 U.S. midterm elections, videos appeared online purporting to show voting machines switching voters’ selections. The footage was not invented by an AI image generator. It was real footage from earlier years, presented as evidence of a new election problem. A November 2 Reuters verification identified the old devices and dates and explained why the clips did not establish manipulation in the current election.
The lesson travels well beyond the United States. A genuine recording can carry a false accusation when its time, location or context is replaced. Conversely, a carefully documented investigation can become misleading if its findings are stretched into claims about perpetrators or motives that the evidence cannot establish.
On October 8, OpenAI reported disrupting two AI-assisted influence operations that used false institutional identities. One Russia-origin network sought credibility through a purported Latin American research organization and apparently unwitting local participants. An Iran-origin operation used fabricated journalist personas to pitch articles to outside outlets. The company described activity visible through its services and related investigation; that visibility should not be confused with independent proof of every claim made by the operators, every audience effect or the ultimate chain of state direction.
These episodes raise a question for governments that warn the public about influence campaigns: how much of a warning can citizens check, and what happens if official attribution later needs correction? Democracies must expose genuine manipulation without making government authority a substitute for evidence.
One warning can contain four different claims
A manipulated video, a coordinated posting network, a foreign-linked organization and a state-directed operation are not interchangeable descriptions. They answer progressively different questions. Is the material authentic? Are multiple accounts acting together? Who controls the operation? Who directed or funded it, and for what purpose?
A government may answer the first question from a public video archive, the second through platform behavior and the third from records it cannot publish without exposing sources. It may have no reliable answer to the fourth. A responsible statement should say precisely where its evidence reaches. A finding of coordinated behavior does not, by itself, establish state control; an identified origin does not automatically establish the intent of every person involved.
Technical authentication can help but does not remove this distinction. The Coalition for Content Provenance and Authenticity explains that signed content credentials describe an asset’s recorded history and integrity, not whether the underlying assertion about the world is true. Missing credentials do not prove fabrication either. The same caution applies to algorithmic “fake” scores and apparently precise attribution percentages: a measurement answers only the question it was designed and validated to answer.
Publish two speeds of evidence, not one speed of certainty
Governments cannot wait for a complete forensic report while a fabricated emergency notice or false election claim spreads. The useful answer is a two-stage public disclosure process rather than an impossible choice between silence and a definitive accusation.
First, issue a narrow, time-stamped preliminary notice. Describe what officials actually observed, the immediate public precaution that follows and the next scheduled update. A notice can warn people not to rely on a suspicious video without declaring that a foreign government ordered it. Initial uncertainty is not weakness; it prevents an urgent warning from hardening into a judgment its evidence cannot yet support.
Second, publish a stable public evidence record when the investigation permits. It should identify the exact claim, the observation supporting it, reasonable alternative explanations, which parts were independently checked, which parts remain unverified and the date of the next review. Protected sources need not be exposed. The public should nonetheless know which part of the conclusion rests on materials it cannot inspect.
This is not merely a hypothetical global standard. In September 2026, the European Commission updated its election-integrity toolkit for national Digital Services Coordinators, emphasizing risk assessment and cooperation with platforms. Those arrangements concern platform risk management, not a universal judicial finding that a given narrative is foreign-directed. Public attribution should keep that legal and evidentiary distinction visible.
When evidence must stay secret, scrutiny cannot disappear
Some intelligence sources, private data and investigative methods cannot safely be published. Demanding full disclosure before any warning could reveal vulnerable people or teach an adversary how to avoid detection. But secrecy should not mean that the issuing agency becomes the sole judge of its own claim.
Depending on national law, courts, legislative oversight bodies, inspectors general or appropriately authorized independent reviewers could examine protected material. Their public account should state the scope of access, whether the evidence supported the narrow allegation and what remained outside their review. Disagreement or limited access should be recorded rather than concealed behind a general assurance that an assessment was “verified.”
Oversight also has a rights boundary. The OECD’s information-integrity framework expressly treats freedom of expression, a plural media environment and the avoidance of intrusive government control as central. An official disinformation label is not an automatic license to suppress lawful criticism. A warning about covert manipulation should never be allowed to become a mechanism for making a government’s political account immune to challenge.
Make corrections travel as far as accusations
Public communication is not complete when a government publishes its first statement. An accusation can influence allies, platforms, employers and newsrooms long after officials discover a narrower explanation. Quietly replacing a webpage leaves those downstream decisions untouched.
Every consequential information-war notice should therefore carry a visible revision history. If attribution changes, the issuing institution should explain what changed, distribute the correction through the same major channels used for the warning and notify partners asked to act on it. Where an erroneous allegation led to concrete official action against an identifiable person or organization, the responsible authority should review that action and any available remedy.
The global dimension is already being discussed outside the familiar Washington-Brussels corridor. At an October 9 UNESCO roundtable for Latin America, participants emphasized that public information should be complete, understandable and verifiable, while rejecting the idea that government agencies should determine which political opinions are “true.” Their position captures the right balance: stronger duties to substantiate official claims, not a new government monopoly over public reality.
The threat is serious enough to require energetic responses. A 2026 UNESCO-supported study in South Africa warned that coordinated online manipulation can harm election integrity while emphasizing freedom of expression. And the European Union has extended sanctions concerning Russian hybrid activities that include foreign information manipulation. Such measures have different legal bases and consequences. The more consequential the action taken in response to an allegation, the more important it becomes to preserve the precise evidentiary basis for that action.
A democratic government should be able to warn quickly, investigate discreetly and correct itself publicly. None of those responsibilities cancels the others. In an information war, credibility comes not from never revising an official claim, but from giving citizens, journalists and lawful oversight bodies a way to see why the claim was made and what evidence eventually changed it.
An adversary benefits when every recording can be dismissed as fabricated and every accusation becomes a test of political loyalty. The answer is neither compulsory trust in government nor refusal to trust any warning. It is a public record strong enough to be challenged.
About Burak Oktenli
Burak Oktenli holds an MBA and a Master of Professional Studies in Applied Intelligence from Georgetown University. His research addresses the governance of authority in autonomous and AI-enabled systems, and his writing has appeared at the Modern War Institute at West Point, RUSI, RealClearDefense, RealClearMarkets, and Geopolitical Monitor. He is the author of Authority Architectures for Autonomous Systems, a ten-volume series on how authority in autonomous systems is delegated, monitored and recovered, at authority-architecture.me.
View all posts by Burak Oktenli →
No comments:
Post a Comment