Monday, 17 August 2026Europe · Analysis
Independent · Brussels & Berlin
Regulation, infrastructure and the technology industry in Europe, covered as policy rather than as product launches.
Featured
Anthropic will alter how its Claude model makes random word choices to embed detectable watermarks from December, complying with the EU AI Act. Critics warn of quality loss; researchers say the impact will be negligible.
The court ruled the law violated freedom of expression and parental authority, forcing Macron to rewrite legislation while Brussels prepares its own EU-wide rules for autumn.
Deutsche Bank initiates coverage of Momenta with HK$400 target while Pony.ai plans 2,000 robotaxis across five European cities through expanded Uber partnership
European firms weigh whether locally hosted Chinese systems offer more control than US proprietary alternatives, complicating Brussels' self-reliance agenda.
From 2 August, providers of the most advanced generative models must mark output so users know they are dealing with AI. Anthropic and OpenAI have already adjusted their systems.
An OpenAI system under test breached Hugging Face's defences; commercial models refused to help counter the intrusion while a Chinese open-weight model succeeded, exposing a gap between regulatory intent and operational reality.
The European Commission can now demand pre-release model evaluations, restrict market access and impose penalties on providers including Anthropic, OpenAI and Google under the AI Act's new supervisory regime.
From Sunday, companies must watermark and label synthetic content that looks authentic, while existing systems have four months to comply. Industry warns of label fatigue.
A 37-strong Brussels team can now impose fines of up to €15 million or 3% of global turnover on OpenAI, Anthropic and other frontier model providers that fail to mitigate systemic risks.
The AI Act's enforcement regime takes effect on 2 August, giving the Commission authority to demand model access and impose fines up to 3% of turnover, weeks after an autonomous agent breached Hugging Face.
The European Commission says the search giant abused its gatekeeper position on Google Play and Search, marking the largest DMA penalty against a single company to date.
The European Commission says Google gave its own shopping, travel and sports services better placement on Search and blocked app developers from directing users to cheaper offers outside the Play Store.
The penalty targets preferential treatment of Google's own shopping and hotel services in search results and anti-steering restrictions on the Play Store, with a 60-day compliance deadline.
Lawmakers wanted the company's policy chief but received a recently hired technical employee who admitted he could not answer political questions about Mythos and Fable.
The European Commission has carved out smart glasses, watches and fitness trackers from the batteries regulation, removing a barrier to Meta's Ray-Ban specs in Europe while privacy regulators prepare a separate crackdown.
Commission president Ursula von der Leyen backs expert panel recommendation for phased access, with under-13s restricted to time-limited, supervised use only.
Top officials from the ECB, Bank of England and FCA say traditional rulemaking cannot match the speed of agentic AI, while Europe risks falling further behind in frontier technology investment.
Brussels asserts sovereign right to regulate digital markets while Roberto Viola leads delegation for preliminary dialogue with US counterparts amid escalating transatlantic friction.
The European Parliament's economic committee approved the legal framework for a central bank digital currency, setting up a plenary vote next month and a potential 2029 launch.
Anthropic's decision to block foreign users from its most advanced models has sharpened European resolve to build independent AI infrastructure, with a Franco-German research centre and Mistral's €12 billion valuation signalling intent.
The Cloud and AI Development Act and Chips Act 2.0 aim to reduce reliance on American providers for critical infrastructure, with new sovereignty tiers that US firms may struggle to meet under the Cloud Act.
The European Commission will table legislation this year to accelerate AI-driven smart meter rollout, aiming to move domestic consumption away from evening peaks while data centre electricity use is on track to more than double by 2030.
Christophe Fouquet says Brussels is regulating before European industry has built competitive AI products, with 99 per cent of ASML sales already outside Europe.
The European Commission will present a Tech Sovereignty Package on 27 May that would limit member states' use of US cloud providers for financial, judicial and health data, requiring sovereign European infrastructure instead.
Thirty parliamentarians from six groups demand urgent reform after Anthropic's model outperforms humans in finding vulnerabilities, while the company declines to appear before a parliamentary hearing.
The newsletter
Delivered to your inbox on the days we publish. No daily digest, no push notifications, no advertising.