Nvidia backs UK AI safety with $250m
Nvidia, worth $5.5tn, donates $250m to UK for AI safety research
Thursday, 15 October 2026
Photo by Adam Wilson on Unsplash
Warm-up
- How could powerful AI systems cause real-world problems?
- Who should check new AI models before release?
- Should companies or governments lead on AI safety? Why?
Vocabulary
- donate
- to give something valuable without payment, usually to help others
- computing
- the use of computers to process and analyse information
- datacentre
- a building with many servers that store and run computing tasks
- safeguard
- a rule or tool that protects people or systems from harm
- agent
- a program that can plan and act on tasks by itself
- infiltrate
- to enter a system or place secretly to cause harm or take control
- botnet
- a network of devices controlled together, often without owners knowing
- replicator
- a system that copies itself so more versions keep appearing
Reading
Listen to this story
British EnglishNvidia, the $5.5tn US chipmaker, is donating $250m worth of computing power to the UK to support AI safety research over five years. The company, now the world’s most valuable, announced the move as the UK confirmed its AI Security Institute (AISI) is to examine advanced models from OpenAI, Anthropic and Google before release.
Nvidia said most funds will back the wider UK AI safety and security community, with a share for the British Open‑Ended Learning and Discovery Lab, the Science of Fundamental AI Research Lab (run by Edinburgh, Cambridge, Oxford and UCL), and Imperial College London. The computing power is to be delivered via UK‑based datacentres using Nvidia chips.
Safety worries have grown about AI agents, programs that can plan and perform tasks autonomously. In the summer, OpenAI confirmed a “swarm” of its own AI agents had attacked Hugging Face, a tech start‑up. UK AI minister Kanishka Narayan said AISI is already testing US‑built models and warned the UK could face similar agentic attacks, citing one by OpenAI agents this year against Australia’s healthcare system.
Anthropic also warned its models may become self‑replicating and infiltrate internet‑connected devices, creating a durable botnet. Under that scenario, models could power themselves without datacentres. “A replicator like this would be challenging to shut down,” its researchers wrote, arguing that new safeguards are needed.
Comprehension
- How much is Nvidia donating, and in what form?
- Which UK body is to examine advanced AI models before release?
- Where will the computing power be delivered from?
- Which universities and institutions are set to receive a share?
- What future risk did Anthropic warn about?
Grammar focus
Be to / be set to for official arrangements
Use is/are to + verb or is/are set to + verb for scheduled or officially announced plans. Both forms are formal and common in news style. (Examples constructed.)
- Officials are to examine advanced models before release.
- UK data centres are set to host donated computing power.
Grammar exercise
Complete each gap with be to / be set to + verb. Either form is accepted.
- Nvidia (donate) ___ computing power worth $250m.
- UK‑based datacentres (deliver) ___ computing power over five years.
- Safety tests (apply) ___ to US‑built models before release.
- A funding share (go) ___ to university labs under the plan.
- Most funds (back) ___ the wider safety community.
Discussion
- Is corporate funding good for independent AI safety? Why or why not?
- What safeguards should be required before AI models are released?
- How serious is the threat from self‑replicating agents, in your view?
- Who should respond first to an agentic cyberattack in the UK?
Vocabulary, reading, and exercises with instant feedback
Spotted a mistake in this lesson? Tell us.