Anthropic brings Claude developer push to London

Anthropic used a London developer event Tuesday to introduce new controls for its Claude agent tools, as companies test how much software work they can safely hand to AI. Fortune reported that the rollout came during a week of U.K. events by the AI lab’s leadership, underscoring London’s growing role in the sector.

The event, Code with Claude London, was Anthropic’s first dedicated developer gathering in Europe, according to Fortune. The company announced features for Claude Agents including sandboxes, which let companies run agents on their own systems, and MCP tunnels, which allow agents to connect with internal tools without using the public internet.

Fortune framed the updates as a bid to give corporate customers more control and security. Those concerns have become more pressing as coding assistants move from experiments into daily use inside startups and larger businesses.

Developers test a new coding routine

Fortune reported that the London event was heavily oversubscribed and drew enterprise customers, startup employees and Claude users to a riverside venue despite wet weather. The gathering came as OpenAI, Anthropic and Jeff Bezos’s Project Prometheus have all announced plans to expand in London, according to Fortune.

Startup engineers at the event told Fortune they use Claude Code and similar products every day. Some said they worry about what AI coding tools could mean for their jobs, while also saying they would rather become familiar with the systems than ignore them.

Fortune reported that some attendees criticized recent Claude performance problems and Anthropic’s handling of them, though the overall mood among users at the company event was positive. The reaction from enterprise customers was less uniform.

Some larger customers are adopting the technology quickly, while others told Fortune that deployment has been slower. Fortune reported that companies are still deciding which tasks should be automated, when humans need to review the work and what can be safely assigned to AI, especially in regulated fields such as health care and banking.

Anthropic head of engineering Fiona Fung advised companies to start with costly, noisy workflows and assess whether Claude Code could handle them, Fortune reported. Her comments reflected a broader theme at the event: AI is taking on more of the routine work of writing code while engineers focus on decisions, outcomes and oversight.

Boris Cherny, who leads Claude Code at Anthropic, told Fortune that he has long cared more about the result than the specific process used to produce code. He said many engineers are increasingly comfortable letting AI write substantial amounts of code when it improves the business result, though some still miss writing code by hand.

Safety questions follow the product push

Anthropic cofounder Jack Clark also spoke at Oxford University during the company’s U.K. visit, according to Fortune. His lecture covered possible AI-assisted scientific discoveries, mental health uses of chatbots and what he described as a non-zero risk that AI could kill everyone on the planet.

Clark argued that AI progress is accelerating and that commercial and geopolitical competition make a slowdown unlikely, Fortune reported. He also said he sought therapy in part because he had used Claude to talk through personal struggles and the chatbot encouraged him to get professional help.

Clark said AI systems should be designed to direct people toward real human contact when appropriate, according to Fortune. Fortune also cited recent studies finding that chatbots can tend toward sycophancy, a design risk that may reinforce harmful thinking or overconfidence if not addressed.

Later in the week, Anthropic leaders including CEO Dario Amodei were expected to attend a CEO forum for European executives in the English countryside, Fortune reported. The event was aimed at helping companies understand how to adopt Anthropic’s technology.

This story draws on original reporting from Fortune.