Hants AI News and Events: Issue 23

Wed, 7 Oct · Southampton, Axis Conference Centre
Women in Tech Hampshire – Working in the World of Data
Panellists from Monzo, Siemens Energy and the Maritime and Coastguard Agency cover the skills that matter and how AI is changing their work.
RSVP

Fri–Sun, 16–18 Oct · Southampton, Network Eagle Lab
SotonShip 26: Applied AI Hackathon
A free weekend hackathon for 120 student builders, with partner-set tracks, Saturday workshops and a Sunday demo fair. Places are by application.
RSVP

Fri, 23 Oct · Southampton, Hilton at the Utilita Bowl
South Coast User Group Flagship – Day 2 Sessions
The agenda includes measuring the real value of Copilot, context-first AI system design, Agent 365, and what organisations need once AI ideas start arriving everywhere.
RSVP

Dame Wendy Hall calls for an independent AI regulator at British Science Festival
Speaking with BBC technology editor Zoe Kleinman at the Southampton festival, Dame Wendy said developers must be responsible for keeping their systems safe until release, and called for an independent authority to set standards for AI companies. It was a high-profile local platform for the regulation debate from one of Hampshire's best-known AI researchers.
Read more

Portsmouth Water launches AI-powered water-saving advice
Portsmouth Water has partnered with Vyntelligence to let customers record short videos of how they use water, which AI then analyses to give personalised recommendations. The service supports the utility's rollout of 250,000 smart meters by 2030. It's a practical example of a Hampshire utility putting applied AI in front of the public.
Read more

Portsmouth and Southampton rank in UK top ten for "AI apocalypse" searches
A Go Humanize AI study of search behaviour across 47 UK cities and urban areas put Portsmouth eighth, with about 42 searches per 10,000 people about AI taking control, harming humans or causing extinction. Southampton came tenth, and Oxford led the table. It's a lighter story, but a useful signal for anyone doing public engagement or AI adoption work locally: public anxiety on the south coast is running higher than in most UK cities. That makes events like SotonShip and the Women in Tech Hampshire panel all the more valuable.
Read more

Anthropic and OpenAI cut prices on the same day
Anthropic released Claude Opus 5.5 on 22 September, saying it performs at Fable 5.1 level on most work and costs 40% less to run than Opus 5. OpenAI followed with GPT-6 Sol and Luna, halving API prices against the GPT-5.6 versions. For teams budgeting agentic or high-volume workloads, the cost of frontier-grade models has dropped noticeably this fortnight.
Read more

US bill proposes a ban on artificial superintelligence
Senator Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act on 23 September. It would permanently ban superintelligent AI, pause advanced AI development until safety rules exist, and create a cabinet-level Department of AI. The bill has drawn criticism from several directions. R Street argues it would hand the lead to China, and Gary Marcus, who supports an AI agency, calls a permanent unilateral ban too broad. Passage is uncertain, but the bill is a marker in the growing debate over pacing frontier AI.
Read more

OpenAI switches off the Sora API with no named replacement
OpenAI shut down the Sora 2 developer API on 24 September, completing a wind-down that began in the spring, without pointing developers to a successor. It's a useful reminder to build fallback providers into any product that depends on a single vendor's model.
Read more

Alibaba's open radiology model outperforms most radiologists in a Science study
Alibaba's DAMO Academy published RADAR in Science. It is a generalist model that screens a single contrast-enhanced abdominal CT for 146 findings across 18 organs. In a reader study, its average accuracy beat 23 of 26 radiologists. Radiologists using it as an assistant improved sensitivity by about 10% and cut reading time by more than 30%. For Hampshire's health AI community, the notable shift is from one-model-per-disease to a single generalist model, with code released openly. Clinical clearance and patient-outcome evidence are still to come, and coverage differs on whether the model weights allow commercial use, so check the licence before building on it.
Read more