Scientists Urge EU Governments to Reject Chat Control Rules

upstart writes:

Scientists urge EU governments to reject Chat Control rules:

As the final vote draws closer, an open letter has highlighted significant risks that remain in the EU's controversial 'Chat Control' regulation.

617 of the world's top scientists, cryptographers and security researchers have released an open letter today (10 September) calling on governments to reject the upcoming final vote on the EU's 'Chat Control' legislation.

The group of scientists and researchers - hailing from 35 countries and including the likes of AI expert Dr Abeba Birhane - has warned that the EU's proposed legislation targeting online child sexual abuse material (CSAM), known colloquially as Chat Control, would undermine the region's digital security and privacy protections and "endangers the digital safety of our society in Europe and beyond".

The group also warned that the new rules will create "unprecedented capabilities" for surveillance, control and censorship, and has an "inherent risk for function creep and abuse by less democratic regimes".

This is not the first time this collective has warned against the regulation, having previously published its recommendations in July 2023, May 2024 and September 2024.

The proposed legislation would require providers of messaging services such as WhatsApp, Signal, Instagram, email and more to scan its users' private digital communications and chats for CSAM material. This scanning would even apply to end-to-end encrypted communications, regardless of a provider's own security protections.

Any content flagged as potential CSAM material by the scanning algorithms would then be automatically reported to authorities.

Currently, 15 EU member states have issued support for the legislation - including Ireland. Six member states oppose the rules, while six remain undecided in their stance.

While the latest draft of the legislation has been amended to exclude the detection of audio and text communications - limiting detection to "visual content", such as images and URLs - the scientists argue that the legislation in its current form is still unacceptable.

The group argues that none of the legislation's changes address its major concerns, namely the infeasibility of scanning hundreds of millions of users for CSAM content with appropriate accuracy, the undermining of end-to-end encryption protections and the heightened privacy risks to EU citizens.

While the latest draft of the regulation has reduced the scope of targeted material (limited to visual content and URLs), the group of scientists states that this reduction will not improve effectiveness.

"There is no scientific basis to argue that detection technology would work any better on images than on text," reads the letter, with further assertions that CSAM detection methods can be easily evaded. The group states that just changing a few bits in an image is "sufficient to ensure that an image will not trigger state-of-the-art detectors".

The group also criticises the EU's proposal of using AI and machine learning to detect CSAM imagery due to the technology's unreliability.

Read more of this story at SoylentNews.