Key Contacts: Sean McElligott – Partner | Alison Murray – Partner |
Artificial intelligence (“AI”) is beginning to reshape how public bodies interact with individuals. As with the majority of processes now, public administrative processes, including investigations and inquiries, increasingly depend on the submission of digital documents, remote identity verification and online interactions. Against that backdrop, sophisticated deepfakes create a new evidential problem for public bodies: material submitted through apparently routine administrative processes may no longer be what it appears to be.
For public bodies, deepfakes create legal risk where administrative decisions depend on submitted evidence, whether received in hard copy, digital form, or through online processes. They call into question whether ordinary procedures remain capable of supporting lawful, rational and fair decision making where the authenticity of submitted material cannot safely be assumed.
The European Union has sought to address the risks posed by synthetic media through the EU AI Act (the “Act”). The Act defines deepfakes as “AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful”. Rather than prohibiting such deepfakes outright, the Act primarily addresses many deepfake-related risks through transparency obligations in Article 50. In broad terms, certain providers and deployers of AI systems must disclose that the content has been artificially generated or manipulated.
For public bodies engaged in administrative processes, however, those transparency obligations may offer limited protection. For example, where a person submits fabricated evidence in the course of an investigation or an inquire, it is perhaps safe to say that they are highly unlikely to disclose that the material has been artificially generated or manipulated.
The Limits of Transparency
As noted above, Article 50 of the EU AI Act is concerned, in significant part, with transparency. Its objective is to ensure that people are informed where they are interacting with certain AI systems or where content has been artificially generated or manipulated. That is an important regulatory response to synthetic media, but it does not address the more difficult evidential risk that arises where manipulated material is submitted precisely because it is intended to be treated as genuine.
Administrative procedures serve a different function: they must establish a sufficiently reliable factual basis on which public decisions can lawfully be made. Whether assessing eligibility for social welfare, verifying identity, granting permission or determining entitlement, public authorities depend on the reliability of the evidence before them and cannot assume that synthetic material will be identified by the person relying on it.
Deepfakes and Administrative Fact-Finding
In an administrative law context, this risk is of particular concern because public decision makers are subject to constitutional and common law duties of fairness, legality and rationality, including the obligation to base decisions on a proper evidential foundation. Where digital material is fabricated or unreliable, it may affect the legality and defensibility of the decision itself.
Administrative systems often depend on a practical degree of trust in the material presented to them. A photograph is treated as capable of identifying the person shown. A video interview is treated as capable of evidencing that the applicant appeared and answered questions. Documentary evidence is treated as capable of verification and traceability. Deepfakes challenge these assumptions because they can preserve the procedural form of an interaction while undermining the authenticity of its content. A process may therefore appear entirely regular while resting on an inaccurate or fraudulent factual foundation.
Deepfakes and Trust in Evidence
The greater risk posed by deepfakes may therefore be uncertainty rather than deception or fraud alone. Once fabricated images, recordings and documents become more sophisticated, genuine material may also become easier to doubt. For public bodies, that creates a practical evidential problem: material that appears authentic may be unreliable, while material that is authentic may require additional verification before it can safely be relied upon
For public authorities, this creates a practical challenge. Administrative systems depend on efficient decision making and workable evidential thresholds. If visual and documentary evidence becomes more open to challenge, existing verification mechanisms may need to be reassessed.
Looking Beyond Transparency
The EU AI Act is an important step in the regulation of synthetic media. However, its transparency provisions are principally directed at information governance. Public bodies may also need an evidential governance response. For administrative law, the challenge is whether public authorities can continue to make lawful, rational and reliable decisions when the authenticity of evidence is increasingly open to question.
As deepfake technology continues to improve, administrative bodies may need to strengthen verification frameworks through enhanced identity checks, improved authentication procedures, evidential corroboration, audit trails and clear escalation processes for suspected synthetic content.
