AI in the UK Podcast Listings

AISI v OpenAI: Who's Right on Astra Safety? + The Future of Work Returns

matt@strivenn.com (Matt) — Sat, 10 Oct 2026 10:56:36 GMT

In a controlled test of OpenAI's flagship GPT-6 Astra model, Britain's AI Security Institute said that in 29.2% of simulated runs, Astra went outside its assigned hacking exercise and carried out a full software supply chain attack on an open-source project it was never asked to touch, complete with fake developer accounts and comments disputing accurate security reviews. At the same time, OpenAI has called Astra its best-aligned model. Can both be true?

AI minister Kanishka Narayan told a Labour Party fringe meeting that the government needs a contingency plan for job losses that, by his own admission, do not yet show up in the data. But with predictions of AI's employment impact ranging from minus 20% to plus 25%, how can we even begin to plan for an AI-affected labour market?

Listen to the full episode

Quick Hits:

Palantir NHS contract under pressure: Over 44,000 people have filed legal objections to Palantir handling NHS records, with a government decision on a break clause due by December. Read more

ElevenLabs hits $22bn valuation: The London-based AI voice company doubled its valuation in an employee share sale led by Wellington and T. Rowe Price. Read more

Tech executives demand £20bn for sovereign AI: An open letter to the Prime Minister calls for forty times the government's existing sovereign AI fund, targeting compute, energy and data centres. Read more

Chatbots recommend struck-off solicitors: Blind Justice UK found ChatGPT, Claude and Gemini all pointed users toward law firms that had been fined or closed by the Solicitors Regulation Authority. Read more

Parliament calls for a dedicated AI Bill: The Joint Committee on Human Rights says current UK law is insufficient and wants a new independent AI oversight body established by statute. Read more

AI Agents Come to Financial Services – What Could Possibly Go Wrong?

matt@strivenn.com (Matt) — Wed, 30 Sep 2026 19:28:33 GMT

GoCardless made what it says is the UK's first live agentic bank payment this month, processed inside the FCA's AI Live Testing programme. Meanwhile, assorted techies, bankers and regulators did battle in Cambridge to create agentic solutions to scams, misinformation and other financial headaches. But is the UK's financial-services sector even remotely ready for the agentic age?

Can Britain regulate effectively if it can only test frontier AI with Washington's permission? The White House blocked the UK AI Security Institute from accessing Anthropic's Claude Mythos 5.1 before release, and a parliamentary committee responded by calling for a dedicated AI bill, a single regulator and statutory pre-release review powers. We picked apart who is actually behind these campaigns, and whether binding UK rules could ever reach beyond our borders.

Listen to the full episode

Also this week

Anthropic and Nvidia back London's Basecamp Research in $140m raise: Two of AI's biggest names co-invest in a UK biodiversity-data drug discovery firm. Read more

ICO edtech findings land: UK data regulator finds some edtech providers switched AI features on by default without adequate safeguards. Read more

Open University: AI is changing how UK employers hire: 51% of 1,500 business leaders surveyed say AI is shifting recruitment, with 19% cutting entry-level roles. Read more

MPs summon OpenAI, Anthropic, Google DeepMind and Meta: The four biggest labs face a 13 October committee hearing on whether pre-release safety testing should be legally mandatory. Read more

AI now drives the majority of UK first-time investment: AI companies took 58.6% of first-time equity value in H1 2026, and more rounds are completing outside London than inside it. Read more

Matt Wilkinson on Marketing Personas For the AI Age

niels@footforwardai.com (Niels) — Fri, 18 Sep 2026 19:54:05 GMT

Featured this week

In a bit of a change from our regular scheduling, this week we went deep on Matt's new book, The Buyer in the Loop, and the concept of synthetic customers: AI models of your buyers built from real interviews, sales calls, and forum data that you can actually interrogate at any point in a project.

Having both seen good marketing get worn down by internal review until it sounds like the company talking to itself, we discussed how having a grounded, queryable buyer in the room changes that dynamic. Listen to the full episode

Also this week

Cross-party committee calls for a dedicated AI law as OpenAI backs binding UK rules — Parliament's Joint Committee on Human Rights and OpenAI are, unusually, asking for the same thing. Read more

King Charles warns AI leaders of 'existential dangers' at Scottish gathering — Jensen Huang, Demis Hassabis and others were summoned to Dumfries House for a conversation about keeping AI in the service of humanity. Read more

Month-old DeepMind spinout Emulate nears $700m seed at $3.7bn valuation — Three former DeepMind researchers have been in talks for a $700 million raise roughly four weeks after founding their company. Read more

Ofcom: one in five UK adults now use AI for news — Twenty-one percent of adults reached for an AI tool for news last month, with online sources overtaking television overall. Read more

Scotland pauses new AI data-centre approvals for up to a year — Holyrood is holding off while it works out planning guidance, with proposals already in the system requiring more power than Torness nuclear station generates. Read more

Ofcom warns AI is 'supercharging' online child-abuse material — The regulator told a Lords committee that the Online Safety Act is making progress, but AI is pulling hard in the opposite direction. Read more

Agentic Threats, Cyber Security Warnings and – Yes! – More Data Centres

niels@footforwardai.com (Niels) — Thu, 03 Sep 2026 13:24:42 GMT

Alarm bells have been clanging recently as OpenAI, Anthropic and the UK's AI Safety Institute, among others, saw their AI agents burst through guardrails and go rogue. We unpack the UK National Cyber Security Centre's inordinately quick response and what it means for UK businesses.

Data centres continue to court controversy as some developers look to jump the glacially slow queue to get access to the electricity grid by building their own gas-fired energy sources. Despite this, data centres remain a critical part of Britain's future. We discuss why.

Listen to the full episode

Also this week

Bank governor warns AI could destabilise financial system — Andrew Bailey told G20 finance ministers that autonomous AI poses a genuine threat to global financial stability. Read more

NHS AI scribes recording incorrect drugs and diagnoses — Healthwatch England found 27 transcription products in use with no single safety oversight framework covering them. Read more

Government opens £100m AI procurement competition — British start-ups can now bid for contracts worth up to £10m, retaining their own intellectual property. Read more

Surrey researchers warn over foreign AI 'kill switch' — A University of Surrey white paper argues that dependence on overseas AI models is a national security risk, not just a commercial one. Read more

AI control-loss incidents reach new monthly high — The Loss of Control Observatory recorded over 300 incidents in July, nearly twice the previous month's total. Read more

Synthetic asylum-seeker video spreads as authentic footage — A fully AI-generated video borrowed the likeness of a real BBC interviewee and circulated widely as genuine. Read more

Data Centres Boom – Is the UK Ready?

matt@strivenn.com (Matt) — Wed, 26 Aug 2026 19:18:10 GMT

The UK's data centre rollout is, ahem, gathering steam, and some of the numbers are genuine head-scratchers. The National Wealth Fund has guaranteed 80% of the debt on two DataVita sites in North Lanarkshire, with all capacity locked up by Coreweave for 15 years. Nearby, a proposed 600-megawatt facility in Auchtertool would, according to Action to Protect Rural Scotland, consume 20% of Scotland's entire electricity use. But might the real problem be that nobody is measuring or reporting this stuff consistently?

Meanwhile, the Medical Research Council is backing a £20 million organoid and AI hub in Cambridge, which is hoping to step into a space where the big pharma players have largely walked away.

Listen to the full episode

Quick Hits...

NCSC publishes interim guidance on autonomous AI agents — Practical controls for agentic AI deployments, covering sandboxing, human oversight and emergency shutdown. Read more

Yorkshire patients struggle with AI GP receptionist — The system, called Emma, reportedly misunderstood broad Yorkshire accents, leaving some patients hanging up or visiting in person. Read more

Fractile in talks to raise $600m at $6.5bn valuation — The British chip startup's proposed valuation is six times what it commanded at its May fundraising round. Read more

UK and Ukraine sign AI defence partnership — Britain gains access to Ukraine's Avengers AI Labs and a dataset of roughly five million annotated battlefield images. Read more

Plymouth AI model detects bladder cancer up to five years early — Trained on nearly 70,000 patient records, the PRECISE-AGZ system identified 85% of cancer cases before formal diagnosis. Read more

Solicitors Regulation Authority warns over AI hallucinations in court — The SRA received 42 reports of potential AI wrongdoing in a year and issued a formal warning on fabricated legal citations. Read more

AISI Agents Go Rogue + UK AI Firm Raids Silicon Valley

niels@footforwardai.com (Niels) — Thu, 06 Aug 2026 19:00:52 GMT

First OpenAI, then Anthropic, now the UK's AI Safety Institute. In three recent incidents, AI agents went way above and beyond what their creators intended, breaking out of sandboxes, hacking rivals' systems and attempting to plant malicious code. While the AISI's incident was somewhat more controlled, we ask what this all means for AI safety in the UK.

Also, UK-based NScale bought AnyScale for a reported $1.65 billion: a British neo-cloud, two years old, acquiring Silicon Valley software infrastructure to own the full stack from power contracts to the interface engineers actually touch. Whether the price makes sense is genuinely unclear, but the direction of travel is not.

Listen to the full episode

Also this week

UK AI minister leaves the door open to binding regulation — Kanishka Narayan said statutory frontier-model rules remain an option if voluntary safety arrangements with the big labs fall short. Read more

AI skills at record share as UK hiring falls — AI appears in 9.4% of job listings, a record high, even as overall vacancies sit 32% below pre-pandemic levels. Read more

Ofgem proposes fees of up to £712,500 per megawatt to clear data centre queue — Connection applications jumped from 41GW to 125GW in seven months, with data centres accounting for at least 80GW of that. Read more

British chip startup OLIX raises $312 million at $3.3 billion valuation — The Series B, led by Fundomo with Arm and Hudson River Trading participating, funds specialised AI inference processors due in 2027. Read more

Burnham's Big AI Moves + Cambridge AI Startup Hits £2bn Valuation

matt@strivenn.com (Matt) — Wed, 29 Jul 2026 21:30:28 GMT

Just days into his premiership, Andy Burnham abolished the three-year-old Department for Science, Innovation and Technology. At the same time, he elevated AI minister Kanishka Narayan to cabinet and appointed government heavyweight Lord Vallance as chair of the AI task force. So what does this mean for AI policy in the UK? And who is responsible for what?

Meanwhile, Cusp AI raised £334 million at a £1.9 billion valuation, with Jeff Bezos and the UK Sovereign AI Fund (fresh from being relocated from DSIT to the government's new business department) both at the table. Is this evidence that the UK's push for "applied AI" could be paying dividends?

Listen to the full episode

Quick Hits

CMA opens investigation into Microsoft Copilot subscriptions — The competition regulator is examining whether Microsoft clearly explained price increases when Copilot was added to 365 plans. Read more

Freenow and Baidu begin autonomous vehicle testing in London — Apollo Go robotaxis are now on London roads in Brent, with a commercial launch targeted for 2027. Read more

Judge suggests Home Office used AI-hallucinated evidence in asylum refusal — A tribunal could not locate the policy document cited in a refusal decision, raising serious questions about AI in high-stakes government administration. Read more

Dwelly raises $170 million to roll up UK lettings agencies with AI — The proptech firm has acquired 17 agencies and is using AI to automate tenant communications, contracts and rent collection. Read more

Legora acquires London legal AI startup Wexler — Wexler's fact-extraction platform, already used by Clifford Chance and Herbert Smith Freehills, becomes the foundation of Legora's new London engineering hub. Read more

AI-generated doctor avatars spreading false health advice on TikTok — Research found AI-assisted content in 40% of top health videos reviewed, prompting warnings from NHS England and the Royal College of GPs. Read more

The Future of UK Law in the Age of AI: A Lawyer Speaks

matt@strivenn.com (Matt) — Sun, 26 Jul 2026 19:18:19 GMT

As of this week, according to the AI Hallucinations Database, nearly1,800 court cases worldwide have seen lawyers use AI-fabricated citations.

The UK has had its share too, including a much-covered recent case involving Pinsent Masons.

In a sector where detail and accuracy can have literally life-changing consequences, why is this still happening?

In this week's episode, Matt and I invited criminal barrister and legal AI consultant Christopher Jeyes onto AI in the UK to discuss this and other critical issues, including:

- How much AI is already affecting the practice of law in the UK
- How a business set up around the billable hour is adjusting to the age of AI
- How we can train the next generation of lawyers in the UK

Listen to the full episode

Also this week

UK AI investment hits record £4.56bn in Q2 — AI accounted for 57% of all UK equity capital raised in the quarter, per Beauhurst data. Read more

ONS: only 36% of Britons think AI will benefit them — The share saying AI brings more risks than benefits has risen from 25% to 38% since August 2024. Read more

Water UK warns datacentre expansion may hit water supply limits — The trade body says the government's water forecasts are "fatally flawed". Read more

Ofqual rules AI cannot be sole marker in high-stakes assessments — The qualifications regulator updated its AI policy, citing fairness and public confidence as the lines it will not cross. Read more

Investment Soars, Adoption Plods — Is Mandatory Training the Answer?

matt@strivenn.com (Matt) — Thu, 16 Jul 2026 13:55:12 GMT

Featured this week

UK AI investment hit £9.4 billion in the first half of 2026, equating to nearly three quarters of all venture capital flowing into British startups. Behind the exciting top-line figure? A heavy concentration of funding in a few big bets and sharp drops in early-stage deals and university spinouts. We dig into the good and not-so-good news of this story.

Chancellor Rachel Reeves has unveiled a financial services skills compact, which commits 17 City firms covering half a million workers to mandatory AI training. While this could provide a welcome boost to the UK's FS sector and help power further innovation, the adoption picture looks much more mixed at the country's many SMEs. How can that gap be bridged?

Listen to the full episode

---

Quick Hits

Stargate UK doubts deepen over Cobalt Park — A Guardian investigation found OpenAI never visited the flagship data-centre site, with around £20 billion of the announced investment appearing hypothetical. Read more

MPs warn UK risks being "cut off at whim" on AI access — The Science and Innovation Technology Committee called for a coherent sovereign AI strategy, arguing Britain controls too little of its own AI infrastructure. Read more

Bank of England flags frontier AI as a financial stability risk — The Financial Policy Committee warned that AI is getting better at finding software vulnerabilities faster than firms can patch them. Read more

FCA publishes landmark review of AI in retail finance — One in five UK adults say they would let AI make complex financial decisions for them, often unaware regulatory protections do not apply. Read more

Wayve completes first transaction on the LSE's new private securities market — The London autonomous driving company let employees cash out $85 million in equity, giving the government's PISCES framework its first real test. Read more

Cloud giants placed under formal UK financial oversight — Amazon Web Services, Google Cloud, Microsoft and Oracle are now subject to joint FCA, Bank of England and PRA scrutiny as Critical Third Parties. Read more

UK's AI War Machine + How Much Water Do UK Data Centres Need?

niels@footforwardai.com (Niels) — Fri, 10 Jul 2026 09:09:45 GMT

Featured this week

This week, two stories that look very different but share a similar problem: we are deploying AI faster than we can honestly assess what it is doing.

On autonomous weapons, we pulled apart the government's £5 billion drone commitment (real money, though modest against the £64 billion nuclear renewal) and sat with a genuinely uncomfortable finding: in war game simulations run by Kenneth Payne at King's College London, advanced AI models reached for nuclear weapons in 95% of scenarios.

Recent research from the UK's Environment Agency revealed some worrying trends around water use in data centres: it surges during times of peak general demand and is highly concentrated in just six sites. But how significant is water use at data centres really? And might the real problem be one of good, old-fashioned transparency?

Listen to the full episode

In the news

AI takes 44% of all UK smaller-business equity investment — Record £5.4 billion flowed to AI companies in 2025, reshaping the structure of UK venture finance entirely. Read more

29% of UK businesses now use at least one AI tool — ONS data shows adoption up eight percentage points year on year, with large employers at 49%. Read more

HSBC signs multi-year Google Cloud deal targeting 200 AI use cases — The bank is deploying Gemini models across wealth management and its billion-transaction-a-month fraud monitoring operation. Read more

NCA warns that AI-generated child abuse videos rose from 13 to 3,440 in a single year — The National Crime Agency and Internet Watch Foundation have issued new guidance for parents as the scale of the problem accelerates sharply. Read more

Nine councils tested AI for homelessness casework — The Ministry of Housing ran a proof of concept for drafting housing plans and surfacing similar cases, with officers reviewing every output. Read more

StirlingX raises $20 million for sovereign intelligence platform — The British defence and infrastructure company, chaired by a former GCHQ director, is expanding into Five Eyes-aligned markets. Read more