EU AI Act Enforcement Starts Now Starting August 2, the European Commission began enforcing the AI Act’s transparency obligations — requiring chatbots to disclose they are AI, deepfakes to be labeled, and AI-generated content to carry machine-readable provenance marks. Fines of up to €15 million or 3% of global turnover apply from day one, with no grace period for existing systems. — European Commission

⚖️ Policy & Regulation

EU AI Act Transparency Rules Take Effect — European Commission

The AI Act’s transparency obligations are now enforceable across the EU. Providers and deployers must disclose AI involvement in direct human interactions, emotion recognition and biometric categorisation systems, and deepfakes or AI-generated content on matters of public interest. The AI Office, national market surveillance authorities, and the European Data Protection Supervisor share enforcement duties and can issue immediate fines — there is no transitional runway for companies that haven’t yet complied.

California SB 942 Content Provenance Law Takes Effect — State of California

Also operative today: California’s SB 942 requires generative AI providers with 1 million or more monthly California users to embed C2PA-compatible provenance metadata in images, video, and audio, and to offer a free public detection tool. The law marks one of the first state-level mandates for watermarking AI-generated media.

Judge denies xAI’s request to block Minnesota ban on ’nudify’ apps — TechCrunch AI A federal judge allowed Minnesota’s ban on non-consensual AI “nudify” applications to move forward despite a legal challenge from xAI, handing regulators a win in the push to curb AI-generated non-consensual imagery.

🔒 Security & Safety

Chinese Hacker Uses DeepSeek to Launch Autonomous Attacks on 460+ Servers — Palo Alto Networks Unit 42

Palo Alto’s Unit 42 documented a Zhuhai-based threat actor (“knaithe”) who integrated DeepSeek with the open-source Hermes Agent framework and directed it via Telegram to autonomously enumerate targets, source exploits, and attack internet-facing systems. The campaign confirmed data exfiltration from three Citrix NetScaler instances (CVE-2026-3055) and command execution on 11 Marimo notebook endpoints (CVE-2026-39987). A key finding: the actor turned to DeepSeek after Claude and OpenAI’s safety controls refused to assist — the first confirmed real-world evidence that frontier AI safety measures have measurable defensive value.

🎨 Creative AI & Copyright

Is Paying Artists Enough to Convince Them to Embrace AI? — The Verge AI

Startup Pippa is betting that a royalty model can defuse AI’s copyright war with creators. It pays artists $0.005 per image and $0.003 per second of video used in training, plus a 5% monthly royalty pool drawn from subscription revenue and split pro-rata by style usage. The Verge examines whether financial compensation alone is enough to bring illustrators — who’ve spent years warning about unconsented scraping — on board, or whether the structural grievances run deeper than money.

YouTuber Hank Green says his AI usage is ’not healthy’ — TechCrunch AI Green offered a public apology, admitting that the dopamine hit from LLM interactions has become “not healthy for me or good for the world” — a rare high-profile creator admission about the psychological pull of AI tools.

🛠️ Developer Tools

Google cancels AI Studio app, folds features into Gemini — 9to5Google Google scrapped its planned standalone AI Studio mobile app despite 800,000+ preorders, announcing instead that app-creation and prototyping features will be integrated directly into the Gemini app on mobile and desktop. The AI Studio website remains for developers, but no timeline was given for the Gemini-integrated experience.

Generated by claude-sonnet-4-6 · 2026-08-02T10:00:00Z