Meta refuses to sign EU’s AI Code

👋 Good morning/evening (wherever you are). It’s Friday.

Earlier this month, Meta refused to sign the EU’s voluntary AI Code of Conduct. It was one of the few major players to say no.

In contrast, Microsoft told Reuters it plans to sign the agreement.

This voluntary code is an early step ahead of the EU’s binding AI Act, which already passed and becomes enforceable in 2025. The goal is to encourage companies to align early and show good faith.

Under the AI Act, companies like Meta, OpenAI, Google, Anthropic, and Mistral will face legal obligations, including model evaluations, risk controls, copyright transparency, and detailed documentation.

Full enforcement begins in August 2025, with fines of up to 7 percent of global revenue.

OK let’s keep going ↓

🔥 Creative pulse

Recap of my top 3 LinkedIn posts this week:

1. Even AI haters will admit this is impressive
   Plenty of honest opinions in the comments. Always fun when LinkedIn stays (somewhat) civilized.

2. Why AI motion capture is bigger than you think
   I break down why this trend matters and what’s next!

3. 24 prompts, 1 fake smartwatch brand
   Every day, it’s getting easier to create spec ads.

💡 Here’s what you should know

• Meta hires two Apple AI researchers

  Meta added Apple researchers Mark Lee and Tom Gunter to its Superintelligence Labs team as part of its ongoing push to attract top AI talent for advanced model development.

• California adopts first statewide court policy on AI use

  The California Judicial Council now requires each court to draft AI regulations by Sept 1, making it the largest U.S. court system to formally govern generative AI.

• ChatGPT may add watermarks to AI images

  OpenAI is testing image watermarking in ChatGPT, with a potential option to save unmarked images, possibly limited to paid users.

• Invideo AI builds full video production on OpenAI models

  Powered by GPT‑4.1, gpt‑image‑1, and text-to-speech, Invideo AI lets users create high-quality videos from simple prompts, automating every step of production.

• Google embeds hotel ads in Gemini AI search results

  Gemini now features paid hotel listings directly within AI-generated travel itineraries, marking a shift in how advertisers reach travelers during early planning stages.

• Perplexity aims to pre-install AI browser on smartphones

  Perplexity is in talks with phone makers to ship its Comet AI browser by default, aiming to break into the mobile search market.

• Oracle adds AI database access with new MCP Server

  Oracle's SQLcl MCP Server lets LLMs securely connect to Oracle Databases, run SQL, and return results, enabling AI agents to perform real-time analysis and workflows with full auditability.

• Citi and Ant launch AI tool to cut FX hedging costs

  Citi and Ant International are piloting an AI-powered solution that helps clients forecast currency risks and reduce foreign exchange hedging costs using Ant’s Falcon model.

• Mizuho taps SoftBank AI to turbocharge banking ops

  Japan’s finance giant is betting on SoftBank’s AI to cut costs, boost speed, and modernize its systems.

• QBE launches AI-specific cyber insurance

  New coverage protects against AI regulation penalties and LLMjacking attacks, including fines, retraining costs, and misuse of model resources.

• Chinese researchers use AI to generate 30,000 climate scenarios

  A new study published in Nature Climate Change shows how deep learning can rapidly create localized mitigation models, helping overcome limitations of traditional climate forecasting.

• AI model transcribes thoughts from brainwaves in new breakthrough

  Scientists at the University of Technology Sydney have developed an AI system that decodes brainwaves into text, advancing the potential for thought-controlled devices without surgery.

• FDA’s AI tool “Elsa” linked to Alzheimer’s treatment potential

  Media outlets spotlight Elsa’s possible role in accelerating Alzheimer’s treatments, though the FDA has not confirmed specific clinical applications.

💰 The numbers

• UK announced a $2.6B plan to expand national AI and supercomputing infrastructure, aiming to boost healthcare, energy, and public services.

• Instacart backer Kris Fredrickson has launched Verified Capital, a $175M fund backing a focused portfolio of AI startups, following past roles at Benchmark, Coatue, and Curology.

• Blaize stock doubled after signing a $120M deal with Starshine to expand its hybrid AI infrastructure across major Asian markets.

• Bedrock Robotics raised $80M (Seed + Series A) to bring autonomous AI systems to heavy construction equipment.

• BrightAI closed a $51M Series A after surpassing $100M in revenue and 250,000 AI endpoints across 50,000 sites.

• CertifID secured $47.5M in Series C funding to expand its AI-powered fraud detection for real estate transactions.

• Firestorm Labs raised a $47M Series A to build AI-driven drone manufacturing systems for defense and commercial use.

• GeologicAI raised a $44M Series B to scale its AI-powered mineral exploration platform, backed by BHP, Rio Tinto, and Breakthrough Energy.

• Benchmark is reportedly leading a $30M Series A for AI code review startup Greptile at a $180M valuation.

• Banyan Ventures launched a $10M seed fund to back early-stage AI infrastructure and workflow startups.

🧠 Thought starters

Are we overvaluing the AI?

Apollo’s chief economist Torsten Sløk shared a striking chart this week comparing today's AI-driven valuations to the dot-com era.

His one-liner: the top 10 companies in the S&P 500 are now more overvalued than they were in the 1990s.

That’s the entire note. Just the chart and a sentence.

😂 Meme of the day

Thanks for reading,

Eddie

Are we friends on LinkedIn and X? Let’s connect — DM me or hit the follow button.

P.S. If this was valuable, forward it to a friend. If you’re that smart friend, subscribe here.

P.P.S. Interested in reaching other ambitious readers like you? To become a sponsor, reply to this email.