SAFE AI NOW Insights - Edition June 2025

Dear Reader,

As we head into summer, AI governance is growing more uncertain. The EU is hinting at enforcement delays, the U.S. is proposing a moratorium, and other countries are pulling in different directions.

But don’t pause your efforts—yet. In this issue, we have summarized the key signals and shared free tools to help you stay ready, act wisely, and plan ahead.

Warm regards,
The SAFE AI NOW Team

AI STANDARDS & REGULATORY UPDATES

🇪🇺 The EU AI Act: Expected delays

Commission discussing a pause on the Act’s 2026 application dates until key standards are ready.

General-purpose model provisions still slated for 2 Aug 2025, but the Code of Practice missed its 2 May deadline; Commission now hints at a staggered launch.

⏩ Link to Politico and Euractiv for further details

🛡️Immediate to do: Don’t bank on a delay—keep readiness plans but build checkpoints so you can slow spend if the pause is confirmed.

🇪🇺 High Risk AI Systems: consultation until July 18 2025.

Public consultation on high-risk AI systems opened 6 June; comments due 18 July 2025. Brussels is seeking real-world examples to refine classification, obligations and Annex III use-cases.

⏩ Link to the Public consultation on high risk systems

🛡️Immediate to do: Review Annex III use-cases; submit comments by 18 Jul; start mapping any HR systems and high risk systems to clauses.

🇺🇸 US - 10-year moratorium

Capitol Hill – Senator Ted Cruz tabled the “AI Regulatory Moratorium Act”, freezing new state & local AI laws for 10 years and pre-empting existing ones. The bill is now before the Senate Commerce Committee.

⏩ More information here

🇬🇧 UK - AI Bill

Government has pushed back its standalone AI Bill by at least a year while it folds copyright and safety into a larger “digital competitiveness” package.

⏩ Access theguardian.com

🇨🇭 Switzerland - Regulatory clarification

On 8 May the FDPIC stressed that the revised Swiss Data-Protection Act (in force since Sep 2023) already governs AI processing, signalling no omnibus AI law is imminent.

⏩ Access the FDPIC communication

🛡️Immediate to do:

Treat AI as regular processing, Update transparency & contract, Be audit-ready

🛠️ FREE TOOLS YOU CAN USE NOW

📫 Practical help is just one click ahead

Looking for tools to evaluate AI maturity, train your team or classify AI systems?
On our website, you’ll find free resources, readiness guides, and tailored tracks for SMEs in regulated industries—all designed to help you build scalable, responsible AI governance.

👉 Visit SAFE AI NOW and take the next step toward compliance with confidence.

The following official, freely available resources help you implement responsible AI governance and align with standards.

Starter Kit for Safety Testing of LLM- Based Applications

Singapore’s free Starter Kit for Safety Testing of LLM-Based Applications (May 2025) packages step-by-step tests for hallucination, toxic content, data leaks and jailbreaks, mapped to EU AI Act & ISO 42001, and plugs into the open-source Moonshot harness,

📥 You can grab it in two clicks:

PDF: “large-language-model-starter-kit.pdf” – it’s the official 115-page guide (look for the IMDA download page). imda.gov.sg
Open-source test harness: on GitHub, go to the AI Verify Foundation repo aiverify-foundation/moonshot; the Starter-Kit tests and README live in the root folder. github.com

🛡️It gives SMEs a zero-budget, immediately usable way to prove their generative-AI controls while larger commercial solutions are still maturing.

Responsible Use of Generative AI

The Responsible Use of Generative AI playbook outlines 10 plays – five for business leaders and five for product managers – that demonstrate how to responsibly use GenAI in day-to-day work and in new products, with accompanying resources and tooling.

📥 Download the Responsible Use of Generative AI Playbook

🛡️Use these playbook to integrate emerging responsible GenAI practices into daily workflows and product development, through actionable steps aligned to strategic priorities.

AI Criteria Catalog For Financial Services

On June 2, 2025, the BSI published a first-ever Test Criteria Catalogue for AI systems in finance.

📥 You can find it on the BSI website here

🛡️How to leverage it?

SMEs can use the criteria to systematically identify and assess AI-related risks in their products and services
Proposes over 100 test items, organized into actionable categories, that SMEs can apply directly or adapt to their resource capacities.
Referencing the EU AI Act: Each criterion maps to specific EU AI Act articles.

Stay connected with AI regulation and ethics updates!

Join our mailing list to receive monthly AI regulatory and ethics updates.
Don't worry, your information will not be shared.

We hate SPAM. We will never sell your information, for any reason.