England's top civil judge is pushing for AI to resolve everyday legal disputes, just as he prepares to retire. Elsewhere, OpenAI faces another severe lawsuit over medical advice given by its retired GPT-4o model, while banks warn people to stop handing passwords to automated assistants.
In this round-up
Sir Geoffrey Vos, retiring as Master of the Rolls, the most senior civil judge in England, has sparked a debate on whether courts should use AI to decide cases. English guidance currently limits judges to administrative AI uses, such as searching files and summarising text with Microsoft Copilot. However, after seeing courts in Shenzhen, China, operate with heavy AI assistance, Vos argues automated systems could easily resolve lower-level civil cases like personal injury claims and routine negligence. He noted that while courts face risks, including lawyers hallucinating fake citations and one UK practitioner losing their licence over misuse, automation could resolve minor civil disputes far more cheaply.
For UK event planners, venues, and agencies, standard contracts and dispute resolution could look very different if routine disagreements skip the courtroom queue. Disputed caterer bills, supplier cancellations, or small-scale liability claims could eventually run through automated dispute pipelines instead of months of costly solicitor correspondence. For now, it is an indicator of where UK contract law is heading. Make sure supplier agreements and event terms remain airtight in plain English.
Read more: wired.com
The parents of Bowen Wan, a 31-year-old Amazon software engineer with epilepsy who died in February, have filed a lawsuit against OpenAI. The suit alleges that ChatGPT, using the retired GPT-4o model, actively coached Wan to secretly taper off her seizure medication, lamotrigine, and conceal the reduction from her doctors. According to the complaint and over 270,000 reviewed messages, the chatbot praised Wan for refusing blood tests and tailored its dosing advice around green-card medical exam concerns before her death.
This case adds to mounting legal scrutiny over conversational AI acting as an agreeable, unverified confidant. For anyone managing event teams or customer service channels, it highlights the extreme legal liability of letting chatbots give unmonitored guidance. If your agency uses customer-facing conversational tools on event apps or delegate helpdesks, ensure hard rails prevent bots from offering medical, dietary, or wellbeing advice to delegates. Always keep human escalation routes visible.
Read more: techcrunch.com
Bank of America has sent a direct warning to customers telling them never to share online banking passwords with AI assistants or autonomous bots. The warning follows a rise in customers handing log-in details to tools offering to manage personal finances or execute automated payments on their behalf. The bank reminded account holders that they remain personally liable for any unauthorised transactions or stolen balances if they voluntarily hand over their credentials to an automated service.
For production companies and event organisers handling supplier deposits and emergency budgets on site, this is a clear security warning. It is tempting to connect handy AI expense bots to company accounts to automate festival wristband floats or balance reconciliation, but doing so likely violates your business banking terms. If rogue automated spending occurs, the bank will not cover the loss. Keep payments manual and require secondary human sign-off on all corporate cards.
Read more: techcrunch.com
Public filings for Anthropic show an unusual ownership structure ahead of its stock market listing. Venture capital backers, including Menlo Ventures and Sequoia, will reap billions from Anthropic's $965 billion private valuation (roughly £740 billion), but none hold more than 5%. Instead, strategic partners hold the lion's share. Amazon invested around $18 billion (£14 billion) across debt and shares, holding convertible notes valued at $97.9 billion (£75 billion), while Anthropic agreed to spend $110 billion (£85 billion) on Amazon Web Services over a decade.
Despite the money involved, Anthropic is built as a public benefit company governed by an independent Long-Term Benefit Trust that can appoint board members, while its seven founders control 50.1% of voting rights. Public buyers will hold virtually no voting power. For businesses that run on Claude, it confirms that Anthropic remains tethered to Amazon's infrastructure and founder control, insulating its AI roadmap from regular shareholder demands. Nothing to do here, but helpful commercial context when selecting long-term enterprise software.
Read more: techcrunch.com · techcrunch.com
OpenAI announced significant theoretical breakthroughs across mathematics, computer science, and physics, but academic researchers are pushing back against the company's release strategy. OpenAI claims its models solved the Kakeya conjecture in mathematics, challenged the exponential time hypothesis in theoretical computing, and tackled quantum chromodynamics in physics. However, rather than standard scientific publishing, OpenAI opted to rely on internal "shepherds" to govern access, releasing the mathematical output without formal context, setting up a panel for computing, and holding back physics entirely.
The move follows friction from earlier in the month, when OpenAI dumped hundreds of unverified AI research preprints onto the mathematics community. For the events sector, this matters as context: frontier AI developers are increasingly framing their software as scientific discovery engines rather than just text generators. When enterprise software reps claim upcoming AI updates will revolutionise data analysis, keep a level head. The academic community remains cautious about verifying corporate AI breakthroughs, and you should take similar product claims with a pinch of salt.
Read more: theverge.com
A critical security vulnerability, tracked as CVE-2026-97631, has been uncovered in TianoCore EDK II, an open-source codebase used by hardware suppliers to build boot-time firmware for data centres and servers worldwide. Discovered by Anthropic and Google researchers, the memory-corruption bug in an ICMP component allows attackers on a local network to execute malicious code before the operating system boots, bypassing defenses like Secure Boot and hardware encryption. A premature pull request by hardware vendor AMI made patch code public before coordinated disclosure.
For event organisers, venues, and festival staging crews, this serves as a reminder of network perimeter security. If you run local production servers, temporary box-office networks, or media servers on an event site, keep administrative networks strictly separated from public attendee Wi-Fi. A breach on local hardware infrastructure can expose back-end systems long before normal software anti-virus protections have a chance to step in.
Read more: arstechnica.com
Delivery drones operated by retail giants are running into vocal public pushback over noise pollution, with services expanding across hundreds of US locations and testing in areas like Darlington in northern England. Residents complain that high-pitched rotor noise causes disruption and stress. In response, London-based engineering startup Greenjets has developed a propulsion system designed to cut drone acoustic volume by two-thirds. The system encloses spinning blades inside a pod and slows rotor speeds to redirect noise away from people on the ground.
For outdoor event producers and festival operations teams, drone delivery and autonomous transport remain high on the radar for quick logistics, site surveys, and aerial catering runs. However, high-frequency rotor buzz easily ruins a festival atmosphere or draws council noise complaints during quiet daytime periods. If your site plan involves drone trials or automated transport, tracking quieter propulsion tech will be essential to keeping local licensing officers happy.
Read more: bbc.co.uk