What happened: A Freedom of Information request found the UK’s Department for Science, Innovation and Technology hasn’t run any trials under its memorandum of understanding with OpenAI — eight months after the partnership launch hype.

Why it matters: This is what ‘AI adoption’ looks like in the wild: memos, MoUs, and optimism, followed by the slow, grinding discovery that public services run on procurement rules, accountability, and the fear of being tomorrow’s scandal. The robots don’t replace bureaucracy; they join it.

Wider context: DSIT pointed to Ministry of Justice ChatGPT use and work with the UK AI Safety Institute, while critics warn voluntary partnerships can dodge normal procurement scrutiny and create lock-in.

Background: The Guardian notes similar UK agreements exist with Anthropic, Google DeepMind, and Nvidia; the article also references broader UK compute ambitions (like ‘Stargate UK’) where timelines and claims are contested.

UK government yet to trial OpenAI tech months after signing partnership — The Guardian

Singularity Soup Take: If you want the real AI story, follow the paperwork. The frontier model race is loud; the actual deployment race is a slow-motion wrestling match between ‘innovation’ and ‘who is liable when this breaks public services.’ Spoiler: liability always wins.

Key Takeaways:

• MoU vs Reality: DSIT said it holds no information on OpenAI trials under the MoU, suggesting the headline partnership hasn’t translated into measurable deployments.
• Procurement Friction: Critics argue voluntary partnerships raise accountability questions because they can sit outside normal procurement processes and performance measurement.
• Lock-In Risk: The piece highlights dependency concerns — once a department standardizes on a vendor’s model and tooling, switching costs and governance constraints harden fast.