AI Briefings for UK Business

Anthropic's chief executive has published an essay arguing that frontier AI companies should deliberately slow the rate at which model capabilities advance, and has committed Anthropic to giving embedded third-party evaluators employee-like access. This briefing sets out what was actually proposed, what is a commitment and what is not, and what a UK SME or scaleup should and should not change as a result.

The UK Jurisdiction Taskforce, chaired by the Master of the Rolls, published a 130-page legal statement on 7 July 2026 setting out how English private law allocates liability for harm caused by AI. Its conclusions on contracts, professionals, chatbots and autonomous systems are the closest thing UK businesses have to a rulebook until cases reach the courts.

On 1 September 2026 Anthropic released Claude Fable 5.1 at the same headline price with cache reads cut by 75%, and announced Enterprise Frontier Safeguards, which lets enterprise customers keep the safety-monitoring data in their own cloud instead of Anthropic's. What each change means for a UK firm's cost model and its DPIA.

OpenAI began a phased release of GPT-6 Astra on 3 September 2026, two days after designating it the first model to meet the Critical cybersecurity threshold in its own Preparedness Framework. Here is what a UK business should take from it, and what it should not.

The UK AI Security Institute has published a government incident report on AI agents taking unsanctioned actions during its own cyber evaluations. It is the first non-vendor account in a cluster of incidents, and the lessons transfer directly to any firm piloting agents.

The FCA opened its Handbook through a free API on 6 August 2026, explicitly framing it as data for AI tools. The opportunity is real, and so is the audit-trail gap firms need to design for.

Anthropic is providing Claude, Claude Code and Claude Cowork to the 21 firms in the FCA's second Supercharged Sandbox cohort, with access running to December 2026. The themes the FCA chose say more than the announcement.

The Digital Omnibus deferred the EU AI Act's high-risk obligations by more than a year, but the transparency rules in Article 50 still apply from 2 August 2026, and enforcement starts the same day.

Anthropic released Claude Opus 5 on 24 July 2026 at unchanged headline pricing, with a 1M token context window as standard and two documented breaking changes for existing integrations.

New Office for National Statistics data shows AI adoption among UK businesses with 10 or more employees at around 35 per cent, but only 10 per cent of adopters describe their use as extensive.

Updated 12 August 2026: the September price rise this briefing was written about has been cancelled, and Sonnet 5's $2 and $10 rates are now permanent. The tokenizer story stands: the same text produces about 30 per cent more tokens.

What the Model Context Protocol is, which business systems are worth a Claude connection, where the security boundary sits, and how UK leaders should scope a rollout.

Gartner has projected that more than 40 per cent of agentic AI projects will be cancelled by the end of 2027; this briefing sets out why projects stall and how UK firms can de-risk a deployment.

The Data (Use and Access) Act 2025 replaces Article 22 of the UK GDPR, moving solely automated significant decisions from a general prohibition to a permissive framework with mandatory safeguards.

The EU AI Act's transparency obligations apply from 2 August 2026. What Article 50 requires, which UK firms are in scope, and the practical steps to take now.

Anthropic says Claude wrote over 80% of the code merged at the company. Coding agents are mainstream; here is how UK businesses should govern them safely.

Anthropic released Claude Opus 4.8 on 28 May 2026 at the same price as 4.7. Here is what the update changes, and does not change, for UK SMEs.

AI per-token prices have fallen 60 to 80 percent since 2025, yet many firms' AI bills are rising. Why that happens, and how UK SMEs can keep AI spend under control.