Anthea Milnes

A large-scale collaborative study by leading AI institutes, led by researchers from @oii.ox.ac.uk, finds many of the tests used to measure the capabilities and safety of large language models (LLMs) lack scientific rigour.

Experts find flaws in hundreds of tests that check AI safety and effectiveness Scientists say almost all have weaknesses in at least one area that can ‘undermine validity of resulting claims’

“Benchmarks underpin nearly all claims about advances in AI,” Andrew Bean @oii.ox.ac.uk told @theguardian.com. “But without shared definitions and sound measurement, it becomes hard to know whether models are genuinely improving or just appearing to.” www.theguardian.com/technology/2...

Social platforms' 'language blind spots' in content moderation bring brand safety concerns A new study of recently mandated transparency data under the EU Digital Services Act found that millions of users of social platforms in the region post in languages without any human moderation.

@themedialeader.bsky.social highlights new insights from @manueltonneau.bsky.social, @deeliu97.bsky.social, Prof. Ralph Schroeder + Prof. @computermacgyver.bsky.social, whose have found that 16mn EU-based X users “do not have moderators for their language.”

uk.themedialeader.com/social-platf...

🎉Many congratulations @manueltonneau.bsky.social @computermacgyver.bsky.social @paul-rottger.bsky.social and co-authors!

Alarmist discourse about the effect of generative AI on elections risks distorting policy, damaging trust, and distracting from the real issues. Read the latest @oii.ox.ac.uk blog post from @felixsimon.bsky.social and @sachaltay.bsky.social

How harmful is GenAI around elections? Will it trigger a misinformation apocalypse and upend elections?

I am happy to finally be able to share @sachaltay.bsky.social & my answers to these and other questions on which we have been working for a year and which is now out via @knightcolumbia.org

Researchers from @oii.ox.ac.uk propose a "backward-walking" approach to risk management for LLMs in medicine, considering potential risks, use cases and trade offs to formulate intended use.

Are we focusing on the wrong things when it comes to US-China AI competition? What is anthropomorphic AI? And is the new Pope a gamer?

The OII community tells all. #OIIBeyondtheBooks @graduate.ox.ac.uk @studyatoxford.bsky.social @ox.ac.uk @socsci.ox.ac.uk

The Oxford Internet Institute will be at #FAccT2024, contributing to key conversations on fairness, accountability and transparency in AI.

Delighted to host our Summer Meeting with researchers from @oii.ox.ac.uk and the Technical University of Munich last week. Fantastic couple of days with over forty researchers taking part in lively discussions and ideas generation. Thankyou for visiting!

Meet Emily and Jenna, MSc students at the Oxford Internet Institute, who share interesting and unexpected facts about their research into life online, and why the OII was their ideal place to pursue postgraduate studies.

#OIIBeyondtheBooks

La destrucción del liderazgo tecnológico de EE UU Las políticas de Trump parecen diseñadas para desmantelar los pilares que han dado al país una gran ventaja competitiva

New! Interesting piece from @carlbfrey.bsky.social @oii.ox.ac.uk writing for @elpais.com discussing Trump’s economic policies and their impact on the US’s tech leadership.
elpais.com/economia/neg...

New! Latest study from @oii.ox.ac.uk researchers Franziska Sofia Hafner, @anavaldi.bsky.social and @rocher.lc shows how AI language models encode a flawed understanding of gender. This poses risks for transgender, nonbinary, and cisgender individuals, particularly with AI used in health tech. 1/5

Why does progress stall—and what can we do about it?

Thrilled to announce that my new book, How Progress Ends: Technology, Innovation, and the Fate of Nations, will be published with @princetonupress.bsky.social on September 16, 2025.

Pre-order it here: hubs.ly/Q03hjRMr0

🚨 New research led by Will Hawkins @oiioxford reveals a surge in publicly available deepfake image generators with nearly 35,000 model variants downloaded almost 15 million times since late 2022. Many are used to create non-consensual explicit images of women.

👾 Gaming can be energising or exhausting, social or isolating, empowering or frustrating. A new study led by Dr Nick Ballou @nballou.bsky.social at @oii.ox.ac.uk shows it all depends on the context.

Will Hungary use facial recognition at Pride and is it legal? | Context by TRF Hungary's threat to use facial recognition at Pride events sets it on collision course with the EU's Artificial Intelligence Act

Great piece @adamndsmith.com @reuters.com
on the (il)legality of using FTF at Pride under the AI Act, incl a quote from me www.context.news/ai/will-hung... @oii.ox.ac.uk

Why do some start-ups thrive while others fail? Dr Fabian Braesemann of the Oxford Internet Institute @oii.ox.ac.uk and Professor Paul X McCarthy @paulxmccarthy.bsky.social explore what the data show about the traits of successful founders in the latest episode of the OII Podcast.

Congrats to the Fairwork programme @towardsfairwork.bsky.social led by Prof Mark Graham of the Oxford Internet Institute @oii.ox.ac.uk - highly commended in Oxford's Social Sciences @socsci.ox.ac.uk Impact Awards for their work to improve conditions for platform economy workers across continents.

‘Kids can bypass anything if they’re clever enough!’ How tech experts keep their children safe online From hacking screen time settings to bypassing website restrictions, young people make responsible parenting in the age of tech feel like a game of whack-a-mole. Here’s how to do it successfully

In yesterday's Guardian, how tech experts keep their kids safe online, including expert comment by @katyahe.bsky.social of the @oii.ox.ac.uk www.theguardian.com/lifeandstyle...

Meta recently announced plans to build the world's longest undersea cable. But what does this mean for digital infrastructure and geopolitics? Professor Vili Lehdonvirta & Anniki Mikelsaar explore the issues in this new blog post from @oii.ox.ac.uk

African Olympiad Academy (Foundation) The African Olympiad Academy aims to develop nascent Math and Science talent from Africa into engaged, globally-competitive problem solvers with a deep commitment to the continent.

Congratulations to @oii.ox.ac.uk alumnus Lofred Madzou and team on the launch of the African Olympiad Academy (AOA). A great initiative set up to nurture the top Math & Science talent from across the continent. All the very best for your new venture! Find out more: bit.ly/4h853co

How do we ensure AI systems are fair, safe and accountable? @swachter.bsky.social and @bmittelstadt.bsky.social of @oii.ox.ac.uk consider the challenges involved in regulating complex technologies and the risks of leaving big decisions to tech giants in the new episode of the OII podcast.

Who are the ghost workers powering AI, and how can we safeguard their rights? Interesting interview with Mark Graham @geoplace.bsky.social of the Oxford Internet Institute @oii.ox.ac.uk

Professor Mariarosaria Taddeo @rosariataddeo.bsky.social steers us through the ethical minefield of AI in defence in the latest OII interview.

Fairwork: Shining a light on the world’s gig economy

Got a big idea that could change the world? Follow it. @geoplace.bsky.social did, with @towardsfairwork.bsky.social, which is elevating the working standards of millions of gig workers worldwide. Mark shares his story here www.socsci.ox.ac.uk/fairwork-shi... #InnovationStories @oiioxford.bsky.social

How is AI reshaping the global labour market? Do we have the skills we need to thrive? Dr Fabian Stephany from the Oxford Internet Institute @oiioxford.bsky.social and El Iza Mohamedou from the OECD - OCDE break it down in the latest episode of the OII Podcast.

🌟 Exciting opportunity for doctoral students! 🌟Researching the social, ethical, or policy dimensions of the internet, digital tech, or data? Share your work, debate ideas, and learn from leading academics on our Summer Doctoral Programme, in partnership with UC Berkeley School of Information.