AI video gets a security upgrade
Plus Ofcom's new chair + latest job moves
Good morning and welcome back. I hope you all had excellent Easter breaks. Yes, I wrote "excellent" because your favourite niche newsletter doesn't do cheap puns...
Necessity is the mother of invention
SPONSORED: Sky-high energy costs, little compute and a lack of
funding — the UK doesn’t have the ingredients to develop its own LLMs right now, but London startup Locai Labs has found an ingenious way to repurpose the best open source models into a UK sovereign LLM. Meet GB1 — it doesn’t train on your conversations, uses licensed data, and is powered by renewable energy. UK fusion pioneers First Light are among the organisations already using it and there’s much more to come.
News In Brief 🩳
Making moves: The government is stepping up its attempts to woo Anthropic, with DSIT officials sketching out options ranging from a London office expansion to a dual-listing when it IPOs, the FT reports. The options will be put to the company's chief executive Dario Amodei when he visits Europe in late May.
See-saw: Ex Channel 4 chair Ian Cheshire is now the favourite to be the next Ofcom chair. Ex-Labour MP Margaret Hodge and former Conservative culture secretary Jeremy Wright are also on the shortlist and were previously the front runners. The FT reckons an announcement could be made as soon as today by tech secretary Liz Kendall. I'm told not to expect it that soon, but the appointment will be made before the end of the month when the current chair's term runs out. The government's preferred candidate will then appear before MPs on the Science, Innovation and Technology Committee. If it is Cheshire there will be plenty of questions about his lack of experience in online safety.
Lots of changes please: The Alan Turing Institute needs to show better value for money, improved governance and a clearer mission, its main funder UKRI found.
Open to ideas: OpenAI has published 20 policy ideas which it believes will build an AI economy that is more open and democratic. It has split them into economic and societal policies and describes them as early ideas to start a broader conversation.
Synthesia's security upgrade
UK AI darling Synthesia announced a string of policy updates this morning to match the latest version of its video platform. The company described it as a “new milestone in enterprise AI assurance”. It comes amid growing pressure in Europe around the use and regulation of AI-generated videos.
The certs: The company has added ISO 27701 — the international privacy standard that covers personal data — to its existing two ISO certifications for information security and AI governance. The UK firm is the first AI video platform globally to achieve all three standards from the Geneva-based International Organization for Standardization.
You know it makes sense: Its head of policy, Alexandru Voica, told us that obtaining the standards would "do more to accelerate safe adoption of AI than any statement of intent or abstract principles". He added: "Our approach also reflects a truth known in the enterprise space: in a fast-evolving world, trust is sustained through evidence and audits.”
Moderate this: Synthesia is also changing how it moderates content to reduce instances of false positives, where legitimate videos are incorrectly flagged for breaching its policies. The London-based company's clients are increasingly using custom avatars based on the likeness of real people, as opposed to stock avatars, based on paid actors or AI-generated characters. But those custom avatars rely on third-party AI models, meaning the company is dependent on their moderation systems and is thus strengthening its own red-teaming efforts.
All you need is trust: Synthesia sells to 90% of the Fortune 100 and Voica wrote in a blog, shared exclusively with The Morning Intelligence, that its updated approach to trust and safety was “designed for enterprise realities", while preventing misuse. In regulated industries like finance, healthcare and government the ISO certifications are independent proof of good practice, making them a sales tool as well as a compliance exercise.
Standards, the way forward: Starling Bank’s chief information officer, Harriet Rees, has also proposed UK standardised testing of AI models used by banks, she told the FT this morning.
Monday Movers (a day late) 👩💻
Get in touch to share your career updates.
On the move: The Treasury's head of AI policy, Diego Duchi, is leaving the civil service after 10 years. Eleanor Boyer, who led on tech investment in North America for the UK government, has relocated to Paris. DSIT's former director of comms Abigail Morris is joining Public First in Washington D.C. as director of strategy and policy.
Hiring: DSIT is looking for a director general for "Digital Foundations". You'll be leading on cybersecurity, digital IDs, digital inclusion and Project Gigabit. Google DeepMind is hiring a policy/comms manager to cover Eve Mannick's maternity leave.
Six days to go: The Morning Intelligence is free for another week until Monday April 13. From that date free subscribers will still receive the newsletter each day, but will only be able to read a preview at the top. Paid subscribers get everything. The first 30 days of all paid subs are free if taken out before April 13. After April 13 only seven days will be free. All revenue goes into producing more independent journalism in this crucial but under-reported area. Prices start from £12.50 a month per person with a group subscription. Do join me if you can.
Thanks for reading, back tomorrow.
Tom