Chancellor warns AI use can destroy legal professional privilege: Sir Colin Birss, Chancellor of the High Court, used a keynote speech at the City of London Law Society on 22 April to warn that AI use across the legal sector carries a real risk of undermining legal professional privilege. The Chancellor made clear that AI does not change the legal test for privilege (legal advice privilege and litigation privilege remain grounded in confidentiality and dominant legal purpose), but stressed that where confidential or legally sensitive material is input into externally hosted or consumer-grade AI systems, privilege may be lost if confidentiality cannot be guaranteed. The speech follows the Upper Tribunal's observations in Munir v SSHD [2026] UKUT 81, which held that uploading confidential documents into "open source AI tools such as ChatGPT" places information "in the public domain" and waives legal privilege. Why it matters for UK lawyers: this is as close to a judicial instruction as practitioners are likely to get before formal rules arrive. Any firm that has not yet drawn a clear line between enterprise-grade AI tools (where data stays within the firm's environment) and consumer-grade tools (where it does not) should treat this as the prompt to do so. The distinction is no longer theoretical. (Judiciary / Local Government Lawyer)

EU AI Act Omnibus trilogue collapses after 12-hour session: the 28 April trilogue on the EU AI Act Omnibus ended without political agreement after 12 hours of negotiation in Strasbourg. The core sticking point was the conformity-assessment architecture for AI in regulated products under Annex I, specifically how the AI Act interacts with existing sectoral safety legislation (the Machinery Regulation, Medical Devices Regulation, and others). Other elements had broadly converged, including postponed deadlines and a prohibition on non-consensual intimate AI-generated imagery. A follow-up trilogue is expected around 13 May. Without an Omnibus passing, the original 2 August 2026 high-risk applicability date legally stands. Why it matters for UK lawyers: any firm advising clients who supply AI-enabled products or services into the EU needs to track this closely. The August deadline is now three months away with no agreed deferral in place. Firms should be advising affected clients to prepare for the original timeline unless and until the Omnibus formally extends it. (IAPP / The Next Web)

Anthropic's Project Deal shows AI agents can negotiate and close real transactions: Anthropic published results from Project Deal, an internal experiment in which AI agents (running on Claude) autonomously negotiated and completed 186 deals worth over $4,000 in a classified marketplace for Anthropic's San Francisco employees. The agents identified potential matches, proposed prices, handled counteroffers, and reached agreement in natural language, without a pre-built negotiation protocol. A performance gap emerged between models: Opus agents completed more deals, extracted $2.68 more per sale on average, and paid $2.45 less per purchase than Haiku agents. Why it matters for UK lawyers: the immediate practical impact is nil (nobody is deploying AI agents to negotiate client transactions today), but the experiment surfaces an important question about what legal frameworks govern AI agents transacting on a human's behalf. Agency law, contract formation, consumer protection, and liability allocation all need answers that do not yet exist. Worth reading for anyone advising on AI governance or commercial contracts involving autonomous systems. (Artificial Lawyer / Anthropic)

Soren launches "private AI" for law firms and regulated industries: Soren, a Y Combinator-backed startup, has launched a service that deploys AI models directly onto a firm's own infrastructure, fine-tuned on the firm's data and governed according to its compliance constraints. The target market is small to mid-sized institutions, including boutique law firms and community banks, that lack the resources to build their own AI stack but cannot accept the data handling risks of sending sensitive information to third-party providers. Why it matters for UK lawyers: data sovereignty and client confidentiality remain the most commonly cited barriers to AI adoption in smaller firms. A product that credibly solves the "where does my data go?" question without requiring an in-house engineering team could shift the conversation for firms that have so far stayed on the sidelines. It is early days, and the product will need to be tested against those claims, but the positioning is sound. (Artificial Lawyer)

82% of UK lawyers use or plan to use AI, but most firms stuck in "automation purgatory": OneAdvanced's Legal Trends Report 2026, published on 28 April, found that four out of five lawyers in the UK are currently using or planning to use generative AI, and nearly half believe they are ahead of their competitors. The reality is less flattering: more than half of firms surveyed report that less than 25% of their work is actually AI-enabled, and nearly two-thirds describe themselves as stuck in what the report calls "automation purgatory" (having made some progress but failing to achieve meaningful transformation). The widening digital skills gap was identified as the most critical barrier. Why it matters for UK lawyers: the gap between AI adoption (installing the tool) and AI integration (changing how work is done) is the defining challenge for most firms right now. This report puts numbers on what many practitioners already suspect: that having access to AI and actually using it to transform practice are very different things. (Intelligent CXO)