22 May 2026: Trump Delays AI Security Order as Google Search Faces Fresh Competition

Trump blinks on AI security rules, Google search faces its biggest challenge in years, and Spotify gives fans a new way to remix their favourite tracks. Here is everything worth knowing from this Friday afternoon.

The US president has delayed signing an executive order that would have required AI companies to submit their models for government security checks before release. Donald Trump told reporters that the language of the draft order “could have been a blocker” and that he did not want to get in the way of the industry, according to TechCrunch. The executive order had been in development for several months and was designed to give national security agencies the power to review advanced AI models before they were made available to the public. Critics argued it risked slowing a sector that moves fast by design. Supporters said frontier models with potential dual-use capabilities needed some form of independent review before wide deployment.

The delay does not kill the policy, but it does reset the clock. White House officials have indicated the language will be revised rather than the effort abandoned entirely. For UK companies working with or procuring from US AI providers, this matters: loosening federal oversight in the United States is likely to accelerate the pace of model releases, but it also shifts more due-diligence responsibility onto buyers. Understanding how AI evaluation actually works has never been more practically relevant for procurement decisions.

Google Search is about to look significantly different, and several competitors are betting that some users will look elsewhere once the changes land. The core shift is Google’s expansion of AI Overviews, which places AI-generated summaries at the top of results pages and reduces the number of website links users see. TechCrunch this week highlighted six search engines worth considering if the new Google experience is not what you are looking for. Perplexity takes an AI-native approach, synthesising answers from multiple sources and citing them, which suits research and fact-checking tasks. Kagi offers a paid, ad-free experience with a strong emphasis on result quality and user privacy. Brave Search, DuckDuckGo, and others have built loyal communities around transparency and data ownership.

The practical question for small businesses and content publishers is not which search engine is best, but what happens to referral traffic when Google answers queries on-page rather than sending users through to websites. Building direct audiences through email newsletters and owned channels is the sensible hedge. If you are evaluating which AI tools to add to your workflow alongside a search engine switch, our guide to the AI tools worth paying for right now covers the current options clearly.

Spotify has struck a deal with Universal Music Group that will allow Premium subscribers to create AI-generated covers and remixes of songs on the platform. Artists whose catalogue is included will receive a share of any revenue the feature generates, TechCrunch reported. The agreement represents a meaningful shift in the music industry’s posture towards generative AI. For most of the past three years, major labels responded to AI music tools almost entirely through litigation and lobbying. This partnership suggests Universal has concluded that participating on favourable commercial terms is more useful than continued resistance. The feature is not yet available in the UK, and Spotify has not confirmed a rollout date for British users. The revenue-sharing model may set a template for how AI-generated fan content is handled in Europe, where copyright questions around AI music remain unresolved.

The global shortage of high-bandwidth memory chips is beginning to push up prices across consumer electronics, and the AI industry’s infrastructure demands are a significant driver. A widely shared piece by writer David Oks argues that the surge in compute needed for AI training and inference is consuming a disproportionate share of global memory supply. The effect is flowing through to the hardware market: mid-range and budget smartphones, tablets, and laptops are becoming more expensive to produce, and some manufacturers are making component trade-offs to protect price points. For UK consumers planning a device purchase in the next six months, the affordable tier is under more pressure than usual. AI development is not contained to software. Its scale-up has real supply-chain consequences that eventually reach everyday products.

Superset, an open-source development environment built specifically for running AI coding agents, launched via Hacker News this week through Y Combinator’s latest cohort. The tool is designed for the increasingly common workflow where a developer is orchestrating several agents in parallel. The founders describe Superset as an IDE for the agents era: a control layer for managing multiple live environments, reviewing outputs, and dispatching new tasks. The project is in early access on GitHub. Whether it finds a substantial audience depends on how quickly coding agent workflows become routine outside the core developer community. That process is accelerating, which makes tools like Superset worth keeping in view even if you are not a software engineer.

Here is everything else worth knowing from today’s AI news.

• Specialisation beats scale: A new Hugging Face blog post argues that smaller, specialist models can outperform large general-purpose models on domain-specific tasks, with direct implications for AI procurement decisions. Hugging Face
• Multi-stream LLMs: A new paper on arXiv proposes running separate parallel streams for prompts, reasoning, and output in language models, which the authors argue could improve speed and reduce errors on complex tasks. arXiv
• Wozniak to graduates: Apple co-founder Steve Wozniak was cheered at a commencement address after telling students that no matter how advanced AI becomes, humans bring something machines cannot replicate. Business Insider
• AI is pushing up smartphone prices: A widely shared blog post argues that AI infrastructure demand is squeezing the global memory supply, pushing up production costs for budget devices and shrinking the affordable tier of the smartphone market. David Oks

The one to watch closely next week is the Trump executive order. The White House has signalled the policy will be revised rather than dropped, meaning a reworked order could arrive with relatively little warning. When it does, compliance timelines for AI providers serving US government clients will likely reset, and UK firms in those supply chains should expect questions they may not yet have answers to.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.