AI Bill no-show prompts UK calls to protect rights | Murdoch: AI is a 'tech breakthrough and threat'
#12 | PLUS: EU AI Act is published | US senators unveil COPIED Act | Perplexity to share ad revs with publishers | OpenAI improves model reasoning | Eno, the first 'generative movie', hits screens
Welcome to this week’s newsletter charting key developments in generative AI and the impacts these are having on human-made media.
🏛️ REGULATION & LEGISLATION
AS WE REPORTED in our Quick Take, a widely trailed AI Bill proved a no-show in the new UK Government’s legislative programme. The King’s Speech said ministers were working on “appropriate legislation” covering “the most powerful AI models”. Shadow tech secretary Andrew Griffith said: “AI safety needs a proportionate approach, not a mystery bill that creates months of uncertainty and loses the UK jobs and investment as it has in the EU.” Lord Clement-Jones, Lib Dem spokesperson for science, innovation and technology in the upper house, told Charting Gen AI the delay was “disappointing”, adding: “We need to clarify IP law in the face of large-scale ingestion of copyright material by generative AI developers and ensure greater transparency.” Conservative peer Lord Holmes — who plans to reintroduce his private member’s AI (Regulation) Bill — said there were “currently significant gaps in the legal framework for governing AI” such as “large language models (LLMs) ingesting creative content without the consent of rights-holders”.
Owen Meredith, CEO of the News Media Association, welcomed the Government’s commitment, telling us: “This is a relatively new and rapidly evolving field and needs appropriate legislation that isn’t rushed. However, without legislation there remains concerns that publishers’ content is being scraped and used by AI and LLM developers without transparency, consent, or reward.” Michelle Stanistreet, general secretary of the National Union of Journalists told us the NUJ would now call for “greater regulation covering the threat of AI to journalists’ and creators’ rights” and lobby for “robust laws which will introduce consent and end the present grand scale theft.” 🔗 King’s Speech; Lord Holmes’ AI (Regulation) Bill
🚨 QUICK TAKE: 'Disappointment' as UK hits pause button on AI Bill
THE FINAL TEXT of the EU AI Act — which takes a risk-based approach to AI regulation — was published. The landmark legislation comes into force on 1 August but various provisions will be phased in over a two-year timeframe. Chapter 5, which deals with general-purpose AI models that power Open AI’s ChatGPT, Google’s Gemini and Anthropic’s Claude, says that from August 2025 the AIs will be expected to share training data details with the European Commission’s AI Office. TechCrunch said it was unclear who would now write the AI Office’s codes of practice while the Financial Times quoted an un-named veteran EU official asking: “What is fair remuneration [for content creators]? What information is protected if it was partly generated by AI?” 🔗 EU AI Act; TechCrunch; Financial Times
THREE US SENATORS introduced a bill seeking to protect “journalists, actors and artists against AI-driven theft, and hold violators accountable for abuses”. Their bipartisan Content Origin Protection and Integrity from Edited and Deepfaked Media (COPIED) Act would direct the National Institute of Standards and Technology to develop standards that would “identify if content has been generated or manipulated by AI” and prevent the removal of watermarks labelling AI content. The bill was widely welcomed by creator groups including the National Music Publishers’ Association, the Recording Industry Association of America and the News/Media Alliance. 🔗 COPIED Act; Senate release
ALSO:
➡ Biden AI adviser “not aware” if White House has a position on “fair use”.
➡ Amazon’s hiring of top execs from AI start-up ‘prompts FTC concerns’.
➡ And Microsoft’s hiring of top Inflection AI staff triggers CMA inquiry.
➡ OpenAI whistleblowers ‘complain to US financial regulator over NDAs’.
➡ Nvidia faces competition authority probe in France, Reuters confirms.
🚨 NEWS MEDIA
RUPERT MURDOCH hailed AI as a “great technological breakthrough” as well as a “new threat” to the news industry. In an interview with Sky News Australia the 93-year-old News Corp and Fox Corporation emeritus chairman said AI would prove a “force for good”. “I don't think there’s any doubt … it will put people out of work. When it does, that will create wealth, and that wealth will be devoted to new industries and new employment,” predicted Murdoch. In May, News Corp signed a content licensing deal with OpenAI, said to be worth $250 million over five years. Under the deal, content from titles including The Times and The Sunday Times, The Wall Street Journal and The New York Post is training OpenAI’s LLMs and will appear in its products. 🔗 Sky News Australia
🌟 KEY TAKEAWAY: Murdoch spoke before the deal with OpenAI was signed and after News Corp CEO Robert Thomson said he preferred “courtship” over “courtroom”. That explains Murdoch’s uncharacteristically conciliatory tone, though at one point he makes clear that if the AIs want to access publisher-owned content “they’re going to have to pay, or they’ll put us out of business”. Time will tell whether $50 million a year for five years represents a fair price for News Corp’s high quality data.
ALSO:
➡ Gen AI poses an existential threat to news, Japanese publishers warn.
➡ Human oversight of AI in newsrooms is “imperative” RISJ study finds.
➡ Traceability of AI-generated content is ‘fundamental for transparency’.
➡ GB News Radio scores UK broadcast first with AI-generated bulletins.
➡ Two 80-something journos are suing OpenAI over ‘systematic pilfering’.
💰 COPYRIGHT & LICENSING
PERPLEXITY, the AI-powered search engine that was last month accused by Forbes of copyright infringement, is poised to launch a revenue-sharing programme with web publishers. Chief biz officer Dmitry Shevelenko told VentureBeat the programme would run in tandem with Perplexity’s advertising drive: “It’s the first of its kind revenue-share programme where if they are contributing a source input for an answer, and we’re monetising that answer with advertising, we’re going share that revenue with those publishers that contributed to that.” The programme is expected to launch soon “with top-tier publishers” as well as prominent bloggers and (wait for it) newsletter authors. 🔗 VentureBeat
🌟 KEY TAKEAWAY: Perplexity hasn’t provided precise details of how the revenue-sharing programme will work. Until it does that publishers have no means of telling whether Perplexity — described by WIRED last month as being a “bullshit machine”, scraping websites without permission and making stuff up — has the potential to quit being a parasite and become a partner. And no, we haven’t been contacted. Yet.
ALSO:
➡ Filmscript summariser Avail to help creators licence content with AIs.
➡ Popular blog brought back from the dead ‘publishes AI-generated slop’.
➡ Australian screen composers want payout for musicians ripped off by AI.
🤖 TECH WATCH
According to Reuters, OpenAI is working on a new approach to generative AI which improves the ability of its models to reason. Reuters said the project had been called Q* but was now known as Strawberry. A source said Strawberry models could not only answer queries but also plan ahead and autonomously conduct “deep research”. Meanwhile, Bloomberg reported that OpenAI had shared a new five-point classification system towards artificial general intelligence (AGI). The AI believes it is currently at Level 1 with conversational chatbots and is on the cusp of Level 2, systems that can solve problems as well as a human educated to doctorate level. 🔗 Reuters; Bloomberg via Yahoo; PYMTS
ALSO:
➡ OpenAI unveils GPT-4o mini, its ‘most cost-efficient small model’.
➡ Scarlett Johansson explains anger towards OpenAI’s Sam Altman.
➡ AI-generated summaries appear in only 7% of Google searches: study.
➡ UK AI firm once considered a rival to Nvidia is acquired by Softbank.
IN BRIEF
🎬 AI de-aging could ‘strip movie stars of their sense of growth as actors’.
📝 Yes, gen AI makes writing easier, ‘but stories end up sounding similar’.
🎮 Game developers say AI will force job changes, ‘and not in a good way’.
💬 QUOTE OF NOTE
“When everybody gets information catered to their needs and interests, packaged in a way assumed to fit their desires, what happens to our shared understanding of events? I believe this is something publishers in pursuit of personal relevance need to pay very close attention to.” — AI in journalism expert Agnes Stenbom, head of IN/LAB at Schibsted, in conversation with the Reuters Institute for the Study of Journalism
🤔 AND FINALLY …
The joy of watching a movie several times is spotting things you didn’t see previously. So imagine the thrill of seeing something different during every viewing. That’s the idea behind Eno, “the first generative feature film”. Director Gary Hustwit devised an AI model that sifts through 30 hours of original interviews with musician and artist Brian Eno and assembles clips with others taken from several hundred hours of never-before-seen archive footage and unreleased music. According to The New York Times there are 52 billion, billion versions of the movie — so every review will need to be different too! Hustwit explains his approach to generative filmmaking in an interview with The Verge. It’s a fascinating read and gets you thinking about the creative potential. Could this be the future of movies? Probably not, but it’s potentially a new art form.
“Charting Gen AI is a very valuable weekly read curating the spinningly rapid AI developments and their impact on media. Graham’s experienced eye in selecting, recognising and commenting on the relevant news in this fast-moving space is a precious source which grants its readers knowledge and acumen with each edition.” — Claudia Vaccarone, Inclusion, Gender & Diversity in Media Advisor