Deepfakes and Procedural Truth: Can We Believe What We See?

The legal maxim "seeing is believing" is dead. The proliferation of accessible and low-cost Deepfake technologies has introduced a critical risk vector in criminal proceedings: the possibility of fabricating audiovisual evidence indistinguishable from reality to incriminate innocent people or discredit testimonies. At Alonso Sala, we have detected an increase in cases where cloned WhatsApp audios or manipulated videos are presented as prosecution evidence in fraud, defamation, or custody dispute cases.

The Forensic Challenge: Digital Noise and Metadata

Detecting a high-quality Deepfake requires going beyond visual analysis. Our defense strategy is based on advanced computer forensic expertise. We do not look for errors in the image (which are becoming fewer), but inconsistencies in the compression level, digital noise patterns (PRNU) that do not match the sensor of the allegedly used camera, and the absence of traceability in the original file metadata.

Deepfakes as a Criminal Weapon

Beyond false evidence, the use of Deepfakes constitutes a crime in itself. We defend victims of:

• CEO Fraud (Deepfake Audio): Use of cloned executive voices to authorize urgent transfers.
• Non-Consensual Pornography (Deepfake Porn): Insertion of the victim's face into adult videos. This is not only a crime against honor, but against moral integrity and, depending on the case, revelation of secrets.

AI Regulation and Criminal Defense

The new European regulation requires "watermarking" on AI-generated content, but criminals do not follow rules. The criminal defense of the immediate future involves demanding algorithmic audits of any digital evidence provided to the process, challenging its authenticity if it cannot be certified that it is free of synthetic manipulation.

State of the Regulation (June 2026)

Update: June 2026. Since this article was first published the framework has moved on, but with no criminal-law changes yet in force: the Spanish Criminal Code (CP) still has no stand-alone deepfake offence. On 26 May 2026 the Council of Ministers approved the draft Artificial Intelligence Act and sent it to Parliament; the bill would amend the CP to criminalise non-consented sexual deepfakes and toughen the penalties for online grooming, alongside administrative fines of up to 35 million euros. While it makes its way through Parliament it is not law, and its content can only be described in the conditional.

Today this conduct is prosecuted under the existing offences: non-consented sharing of intimate images (Art. 197.7 CP — whose fit with purely synthetic images is debated, as the provision requires images obtained with the victim's consent in a home or other private setting), the offence against moral integrity (Art. 173 CP), insult, slander and fraud. At the European level, the EU AI Act's obligations for high-risk systems become fully applicable on 2 August 2026, and its Art. 50 requires the labelling of AI-generated content under AESIA's supervision — one more tool in support of authenticity expert evidence.