Leaked doc: EU considers a ban on using AI for social credit scores, surveillance, and more; non-compliant companies could be fined up to €20M or 4% of turnover
Politico
Context & Ripple Effects
This Politico leak is the draft stage of what became the EU's AI rulebook within a week: the formal proposal that followed regulated "high-risk" uses like loan applications and raised the fine ceiling to 6% of global revenue. The leak also fits an established Brussels playbook — three months earlier, a leaked data-rules document previewed fines of up to 10% of annual revenue for giants like Google and Amazon before those rules were unveiled.
What makes this draft notable is its red-line structure: some uses are not merely risky but banned outright, separating prohibited categories (social credit scoring, surveillance) from the high-risk tier that carries compliance duties.
First-order effects
Companies building or deploying AI for social credit scoring or surveillance face an outright prohibition rather than a compliance burden, with exposure up to €20M or 4% of turnover for non-compliant uses.
Second-order effects
Vendors selling into Europe will need to sort their portfolios into banned, high-risk, and permitted tiers — the proposal's treatment of loan applications as high-risk shows the tier boundary lands on everyday financial products, forcing reclassification and documentation costs across adjacent markets.
Third-order effects
The penalty-ladder approach here hardened over time: by 2024 the Commission was warning Microsoft of fines up to 1% of global turnover over generative-AI risk disclosures, and by 2026 the AI Act's enforcement powers let regulators evaluate models before regional release and restrict market access — turning EU AI law from a fine regime into a gatekeeping one.
The trend: The EU is converting leaked drafts into graduated AI enforcement — from percentage-of-revenue fines toward pre-release evaluation and market-access control.
Some of the AI systems that could be banned in Europe under these rules may include those used for “indiscriminate surveillance applied in a generalised manner” and “social scoring”. https://www.bbc.com/...
There's a fine balance between AI's benefits and harms, but I believe we need clear and consistent guidance, and tougher regulation to ensure the responsible development of these ‘high risk’ technologies. #AI @oiioxford @GPAI_PMIA https://www.bbc.com/...
An interesting effort to define which uses of #AI are useful to,society, and which ones are not. Europe seeks to limit use of AI in society. https://ow.ly/...
European Union is regulating AI to explicitly allow the use of deepfakes. Is this a good idea? It's a risk in policy, politics, and disinformation. People are bad at spotting doctored content. Probably nobody will see the warning signs? https://blog.lukaszolejnik.com/ ...
EU is considering new AI regs to limit or stop “high risk” uses. “[Proposal will] prohibit AI systems that cause harm to people by manipulating their behavior, opinions or decisions; exploit or target people's vulnerabilities; and for mass surveillance.” https://www.politico.eu/.…
Europe seeks to limit use of AI in society - The use of facial recognition for surveillance, or algorithms that manipulate human behaviour, will be banned under proposed EU regulations on artificial intelligence. https://www.bbc.com/...
“Any companies that develop prohibited services, or fail to supply correct information about them, could face fines of up to 4% of their global revenue” An actual step in the right direction for tackling inbuilt bias If your product doesn't work for everyone, it's unfinished http…
The EU's AI laws are the first of their kind. We got out hands on the leak of the regulation, and here's what we know about the EU's “third way” for AI. https://www.politico.eu/...