Beyond True or False: FIMI, AI and the Future of Digital Trust
This publication is written by dr. Andreja Mihailović, Teaching Associate from University of Montenegro, as part of the Ulysseus project METACOG.
What if the most effective form of manipulation does not require a single false statement?
For years, the public conversation around disinformation has been framed largely as a question of accuracy: Is this information true or false? Can the image be verified? Did the event really happen? Is the source credible?
Those questions remain essential. But they are no longer sufficient.
The contemporary information environment increasingly requires us to look beyond individual pieces of content and examine the behaviour surrounding them: who created the message, how it was distributed, whether multiple accounts are acting in coordination, which audiences are being targeted, whether artificial intelligence was used to generate or manipulate the material, and what wider objective the communication may serve.
This is where the concept of Foreign Information Manipulation and Interference (FIMI) becomes particularly important.
The European External Action Service uses FIMI to describe manipulative patterns of behaviour that may threaten political values, procedures and processes. The significance of this approach is that it shifts attention from the narrow question “Is this statement false?” towards a broader one:
How is information being used to influence the way people perceive reality?
Manipulation does not need to invent facts
A sophisticated influence operation does not necessarily depend on fabricated information.
Authentic photographs can be presented outside their original context. A genuine statement can be selectively edited. Old footage can be presented as evidence of a current event. Real social tensions can be strategically magnified. Networks of coordinated accounts can create the appearance that a marginal position represents widespread public opinion.
The individual pieces may contain elements of truth, while the overall informational picture becomes deeply misleading.
Artificial intelligence adds another layer to this problem.
Generative AI systems can produce convincing text, voices, images and video at a scale that was previously expensive and technically demanding. Synthetic personas can communicate continuously. Messages can be translated, adapted and personalised for different audiences. Images and recordings can be altered with increasing precision.
The question of trust therefore becomes more complex than separating authentic content from fabricated content.
We must also understand how authenticity itself can be strategically manipulated.
When anything can be fake, everything can be denied
Deepfakes are usually discussed in terms of false evidence: a person appears to say or do something that never happened.
Yet generative AI creates an opposite risk as well.
Once citizens become accustomed to the idea that convincing audio, video or images can be artificially generated, genuine evidence may also become easier to dismiss as artificial.
This phenomenon is often described as the “liar’s dividend”: the existence of convincing synthetic media gives people an additional opportunity to challenge the authenticity of real material.
Digital distrust can therefore become as damaging as digital deception.
A society in which citizens believe everything they see is highly vulnerable to manipulation. But a society in which citizens believe nothing they see faces a different problem. Public debate, journalism, institutional communication and democratic accountability all depend on some capacity to establish shared facts.
The central challenge is therefore not blind trust.
It is calibrated trust – the ability to assess information according to evidence, provenance, context, source credibility and the methods through which content reaches us.
Europe is beginning to regulate the transparency problem
The EU Artificial Intelligence Act introduces an important regulatory response.
Since 2 August 2026, Article 50 of the AI Act applies transparency obligations to certain AI systems and AI-generated or manipulated content.
Providers of AI systems that generate synthetic audio, image, video or text must ensure that outputs are marked in a machine-readable format and are detectable as artificially generated or manipulated, subject to the conditions established by the Regulation.
The rules also address the use of deepfakes. Deployers of AI systems that generate or manipulate image, audio or video content constituting a deepfake must disclose that the content has been artificially generated or manipulated.
Specific transparency requirements also apply to certain AI-generated or manipulated text published for the purpose of informing the public on matters of public interest, while the Regulation recognises particular exceptions and safeguards, including situations involving human review and editorial responsibility.
These obligations represent an important shift.
Transparency is no longer treated merely as good practice. In defined circumstances, identifying artificial generation or manipulation has become a legal obligation.
The objective is not to stigmatise AI-generated content. Synthetic media can serve legitimate creative, educational and commercial purposes. The deeper regulatory idea is that people should have relevant information about the nature of the content they encounter.
Trust requires context.
But labels cannot solve the problem alone
Even a perfectly labelled AI-generated image can be used manipulatively.
A disclosure tells us something about how content was created. It does not necessarily tell us why it was created, who is distributing it, whether its presentation is deceptive, or whether its dissemination forms part of coordinated activity.
This is where AI transparency and FIMI analysis meet.
Technical detection can identify signals.
Transparency rules can provide information about provenance.
Platform policies can respond to coordinated behaviour.
Journalists and researchers can investigate networks and sources.
But citizens still need the cognitive skills to interpret what they encounter.
That makes AI literacy and critical thinking part of the infrastructure of digital trust.
From detecting fake news to understanding information systems
The next generation of digital literacy cannot be reduced to a checklist for spotting fake headlines.
Citizens need to understand that information can be manipulated through content, context, coordination, amplification, impersonation and artificial generation.
They need to ask:
- Who is the source?
- Can the claim be independently verified?
- What is missing from the context?
- Why am I seeing this content?
- Is the apparent popularity of the message organic?
- Has the material been artificially generated or altered?
- Who benefits if I accept – or reject – this information?
These are precisely the kinds of questions that initiatives such as METACOG can bring into education.
The aim should not be to teach people to distrust technology, media or institutions. Nor should education create the illusion that every user can become a professional fact-checker.
The goal is more demanding: to develop citizens who understand how contemporary information systems work well enough to question them intelligently.
The future of digital trust will not depend on our ability to classify every statement as simply true or false.
It will depend on whether we can understand the systems, technologies and behaviours that shape what appears before us – and still make informed judgments when certainty is difficult.
In the age of AI, critical thinking is no longer only about deciding what to believe. It is about understanding why something was made believable in the first place.