Key points
- AI Act, Art. 50: since August 2, 2026, providers of generative systems must mark content in a machine-readable format, and anyone publishing a deepfake must disclose it.
- Italy: since October 10, 2025, disseminating a deepfake without consent that causes unjust harm has been a criminal offense, punishable by one to five years’ imprisonment.
- For companies: labels on their own content, and procedures against fraud using cloned voices and faces.
A phone call in the CEO’s voice asking for an urgent wire transfer. A video of a politician saying things they never said. A fake intimate image of a real person. Deepfakes are no longer a lab experiment: they are tools of fraud, disinformation and violence. In 2026, the rules to counter them are finally in force, both at European level and in Italy.
What the AI Act requires
Article 50 of the AI Act, applicable from August 2, 2026, splits responsibilities between those who build the systems and those who use them to publish content.
For generative systems already on the market before August 2, 2026, the marking obligation kicks in on December 2, 2026: this is one of the effects of the Digital Omnibus, which also introduced, from the same date, a ban on apps that generate non-consensual intimate images.
Code of Practice and guidelines
To put Article 50 into practice, the Commission facilitated the drafting of a Code of Practice on transparency of AI-generated content, published in its final version on June 10, 2026. It has two sections: one for providers (marking and detection) and one for deployers (labeling of deepfakes and texts). At the end of July 2026, it had around 190 signatories. On July 20, 2026, the official guidelines on transparency obligations were also published.
How to recognize artificial content
No technique is infallible on its own. That is why the emerging approach is a layered one:
Watermarks and metadata can be removed or lost when a file is recompressed or photographed from a screen. The visible label depends on the good faith of the publisher. In the end, the last filter remains the critical sense of the viewer.
Italian law: the deepfake offense
With the Italian AI law (Law No. 132 of September 23, 2025), in force since October 10, 2025, Italy was among the first European countries to introduce a specific criminal offense. The new Article 612-quater of the Italian Criminal Code punishes with one to five years’ imprisonment anyone who causes unjust harm to a person by disseminating, without their consent, images, videos or voices that have been falsified or altered with AI and are capable of misleading as to their authenticity.
- Prosecution: a complaint by the injured party is required, but the offense is prosecuted ex officio if it is connected to another offense prosecutable ex officio, if the victim is incapacitated by age or infirmity, or if the victim is a public authority targeted because of their functions.
- General aggravating circumstance: the use of AI becomes an aggravating circumstance when it is an insidious means, hinders the defense or worsens the consequences of the offense.
- Markets: market manipulation committed with AI is punishable by two to seven years’ imprisonment.
What to do in your company
- Label your own AI content. Promotional videos with avatars, synthetic voices, generated images: an internal policy on when and how to disclose it.
- Protect payments. No urgent wire transfer authorized solely by phone or video call: call back on a known number, require dual approval, use code words for unusual requests.
- Train people. The AI literacy required by the AI Act also includes being able to recognize the signs of manipulation.
- Be ready to respond. A procedure for reporting, removing and refuting a deepfake involving the company or its executives, with legal support for any criminal complaint.
The compass
Transparency does not eliminate deepfakes, but it changes the rules of the game: those who use AI in good faith have a clear way to disclose it, while those who use it to deceive now risk penalties and, in Italy, prison.
This article is for informational purposes only and does not constitute legal advice. Information current as of November 2026.




Leave a Reply