Signal AI Monday

Your weekly beginner-friendly roundup of what's happening in AI. 🚀

🤖 The Big Stories

OpenAI drops GPT-5.4 — Just 48 hours after GPT-5.3, OpenAI released GPT-5.4 with native computer use, meaning it can control your mouse and keyboard to use desktop apps and websites. It scored 83% on professional knowledge tests (matching experts in 44 occupations) and is 33% less likely to hallucinate. Also supports a 1 million token context window.

Anthropic fights back with Claude 4.6 — Two variants: Opus 4.6 for serious reasoning (already found 22 high-severity Firefox bugs), and Sonnet 4.6 for coding. Though they had a rough week — the Pentagon labeled them a "supply chain risk."

Google's Gemini 3.1 Pro dropped with improved multimodal reasoning.

DeepSeek V4 — A 1 trillion parameter model that's 40% more memory efficient and 1.8x faster.

The bigger picture: The AI industry is shifting from chatbots to agents — AI that actually does tasks for you, not just talks.

🇪🇺 European Corner

August 1, 2026 is the date — That's when the EU AI Act's core requirements fully kick in. High-risk AI systems (like credit scoring or transaction monitoring) will need to comply with strict transparency, data governance, and human oversight rules.

What's changing: - Transparency rules for AI-generated content start Aug 2, 2026 - Product Liability Directive kicks in by December 9, 2026 — AI software is now a "product" under EU law - Digital Omnibus proposal aims to simplify some of the messier AI Act rules

Fines? Up to 3% of global turnover for non-compliance.

The UK, by contrast, is taking a more relaxed, pro-innovation approach.

📊 Quick Hits

• OpenAI raised $110 billion in new funding
• Nine major AI model launches in the last 4 weeks
• Frontier AI now achieves 24% real-world effectiveness (vs 90% in benchmarks — there's a gap!)

That's a wrap! See you next Monday. 🧠

— Adam