👋🏻 Hello! We’re now 2,200+ reading ”Sensible AI” to move from headlines to decisions, in minutes.

Today, governance moves forward: the EU is considering adjustments to the AI Act and a code for labeling AI-generated content, Ireland investigates LinkedIn for improper data use, and New Zealand introduces a biometric privacy code.

We’ll close with your “from the news to your action plan” section and a key resource to understand where the datasets used by AI systems come from.

Let’s begin 👇

^{Spanish version here: IA con Sentido – Español}

TOP NEWS

🇪🇺 EU considers easing AI rules
→ A leaked draft suggests the European Commission could extend compliance deadlines for the EU Artificial Intelligence Act to August 2027 and exempt certain systems from registration.

🏆 UNESCO promotes an ethical-AI prize
→ UNESCO and Uzbekistan award the Beruniy Prize to three researchers and one institution for advances in AI ethics and governance in 2025.

🏛️ Queen Elizabeth Prize: a rare gathering of experts
→ King Charles III presented the 2025 QEPrize to Yoshua Bengio, Bill Dally, Geoffrey Hinton, John Hopfield, Jensen Huang, Yann LeCun, and Fei-Fei Li.

WHAT HAPPENED WORLDWIDE?

🇪🇺 European Commission: AI content-labeling code
→ Ongoing work on a “Code of Practice for the marking and labeling of AI-generated content,” with voluntary application starting August 2026.

🏦 Ireland investigates LinkedIn over AI data
→ The Irish Data Protection Commission examines the use of users’ personal data for training generative AI.

🇪🇺 European Commission — AI Act amendments
→ Will propose eliminating the registry for self-exempted systems—meeting civil-society pushback—with Parliament and Member States to decide.

🇺🇾 Uruguay — a balanced AI law
→ The government begins work on legislation to regulate AI and balance power between corporations and citizens (no date set).

🗺 India publishes AI guidelines
→ Announces flexible national guidelines for AI with limits and a focus on responsible regulation.

🇬🇧 United Kingdom limits judicial use of AI
→ Judges may not use chatbots with confidential data and must verify results after 18 false citations in an £89 million lawsuit.

🇳🇿 New Zealand — Biometric Privacy Code
→ In force with 13 rules, it bans biometric categorization and sets a compliance deadline of August 3, 2026.

AND AT ENTERPRISE LEVEL?

🤖 OpenAI — progress & recommendations 
→ Projects systems capable of discoveries in 2026–2028 and urges shared standards and public oversight to maximize benefits and reduce risks.

🔐 EU-Startups — governance momentum 
→ UK’s AI Score secures €864,000 pre-seed, addressing AI governance—a priority cited by 70% of security leaders.

🌱 Hisense pushes AI for sustainability
→ Accelerates ESG strategy by integrating AI for manufacturing and carbon reduction; four carbon-neutral factories certified in China since 2024.

ORGANIZATIONS & STANDARDS

🌐 UNESCO adopts neurotechnology standards
→ Approves a global ethics framework for neurotech—including AI components—with 100+ recommendations on neural data and mental privacy.

🛡 UN presents commission on military AI
→ The UN and the Global Commission discuss a 2025 report on responsible design, risks, and ethical standards for military AI.

FROM THE NEWS TO YOUR ACTION PLAN

I turn these headlines into concrete steps you can start today to strengthen your AI governance. Pick 1 action to drive results under 30 days:

🟪 Corporate AI committee → Create a cross-functional committee with executive leadership to oversee risk, compliance, and ethics in AI projects.
🟪 BYOAI policy & training → Set clear guidelines for external tools; add training modules on privacy, bias, and cybersecurity.
🟪 Data & metadata management → Document sources, licenses, and data transfers per privacy norms like New Zealand’s code or the Ireland case.
🟪 Ethics & human rights → Incorporate UNESCO principles on neurotech, mental privacy, and social impact across the lifecycle.
🟪 External participation & alliances → Join public consultations, academic networks, and international forums (UNESCO, UN, governments) to anticipate new rules.

RECOMMENDED READS

PwC confirms Responsible AI is already delivering value: 58% report improvements in ROI and efficiency, with growing benefits in customer experience and innovation.

The big challenge now is operationalizing governance (moving from principles to measurable processes) and adapting controls to AI, with responsibility increasingly close to technical teams under the 3 lines of defense model.

USEFUL RESOURCES

🔴 The Dataset Explorer

This powerful tool maps the “DNA” of thousands of datasets used to pretrain and fine-tune models (text, voice, and video).

It traces provenance, legal conditions, and composition (including language coverage). This enables compliance (copyright, attribution), bias evaluation, and informed choices about what data to use or exclude.

RESPONSIBLE HUMOR

That’s all for today!

What topic would you like me to dig into on Friday? Your reply guides the next edition.

If this helps your team, forward it and tell me what worked: I improve the newsletter with your feedback 💌