As artificial intelligence (AI) becomes smarter, the creation of “deepfakes” is increasingly a social and legal problem.
What are deepfakes?
Deepfakes are videos, pictures or audio soundbites that have been synthesised or manipulated using generative AI to look real. They are usually malicious or exploitative where they represent prominent people, such as politicians or media personalities. Widespread use may undermine trust in what we read on social media or see or hear in broadcasts. The Prime Minister, Kier Starmer, while still Leader of the Opposition, was reportedly left feeling embarrassed following a deepfake video being posted showing him swearing at staff. More recently, a deepfake appeared to show Taylor Swift endorsing Donald Trump.
Fake images can also be posted with the intention to cause humiliation or distress to a person by the manipulation of intimate images.
Deepfakes have the potential to mislead, misinform and cause reputational damage.
On account of the prevalence of generative AI, it is now easy, even for those with little technical knowledge, to access and manipulate data to create deepfakes.
Are deepfakes against the law?
The difficulty with rapid technological advancement, is that the law struggles to keep up with the pace, and so, while the law around AI is constantly evolving, any gaps in regulation can leave opportunities for potential exploitation. What, if any, laws have been broken by a published deepfake image will depend on the facts and circumstances of the case.
Criminal sanctions
There are criminal law sanctions, but these are not comprehensive and are likely to be reinforced.
Possibly, the most hurtful deepfake uses have been where they have been used to alter or create intimate images which are then published to demean and cause distress to individuals. Section 188 of the Online Safety Act 2023 has inserted the specific criminal offence of sending intimate images without the victim’s consent and with the intent of causing alarm, distress, or humiliation (“revenge porn”) into the Sexual Offences Act 2003. The offence can be committed without fakery but the reference to appearing to show an intimate photograph demonstrates clearly that deepfakes are caught as well. There is the prospect of stronger laws in this area proposed in the Criminal Justice Bill that was at Report stage in the House of Commons before the General Election.
Intellectual property and reputation management
Using existing data or images and adding new content that is not original may constitute an infringement copyright. The original works may be substantially reproduced and so infringe the original copyright.
Copyright law may therefore still be enforceable against identifiable deepfake creators. Stopping harmful deepfakes is however a matter for copyright owners to enforce through the civil courts. There will be possible defences such as that of no copyright existing or being infringed in the first place, or that a deepfake image is protected by the fair dealing defence of parody in section 30A of the Copyright, Designs and Patents Act 1988. A determination that a deepfake is no more than a lawful parody will depend upon the intention of the alleged infringer as well as the content and context of the image.
It would be easy for an unscrupulous advertiser to manipulate a celebrity’s image to make it seem that he or she was endorsing a product. Although, case law has not so far related to deepfakes, celebrities such as Rihanna and Eddy Irvine have obtained damages for passing off where their genuine images have been used without their permission. A recent complaint was made to the Advertising Standards Authority concerning Simmer Limited, an online meals delivery service, which spliced footage from the Dragons’ Den TV show to give the impression that the “Dragons” directly praised their food products.
Using a digitally manipulated image to show a person in a light that invites hatred, ridicule or contempt can, of course, also be considered libellous.
Regulatory responses to deepfakes
Ofcom published a paper entitled Mitigating the Harms of Deceptive Deepfakes on 23 July 2024, identifying deepfakes intended to humiliate individuals, misrepresent identities as part of a scam, or create disinformation by spreading falsehoods, influencing opinions, or creating distrust. It has identified forms of intervention that can restrict or eliminate harmful deepfakes. The document is a general call to action for measures to be taken to reduce the risks, but one which doesn’t propose any further specific legislation.
It will be interesting to see how the recently announced AI Bill will regulate all aspects of generative AI (including deepfakes) to mitigate the various risks that AI poses to society.
For more information on all intellectual property and technology matters, please contact Laurie Heizler – laurie.heizler@twmsolicitors.com.
For further details about our Business Law Services please click here.
