Greetings all, from the peak of the UK’s fourth heatwave this year, insane.
I’m just finished teaching on an Oxford-George Washington Summer School (well, apart from the final dinner tonight, but dinner & drinks are very much within my field of expertise). It was incredibly intense, but it was also incredibly refreshing. I had an amazing group of students, and the faculty are some of the most incredible (and lovely) human rights people you could hope to meet. So while i’m very much looking forward to a few days off, I also feel very recharged. Thank you!
On that note, this will be the last newsletter for a while. I need to focus on some research for the next couple of weeks. And then i’m focusing on time away from work for an as yet undecided time.
On the naughty step this week is the Home Office. [… leaving some time for you to gather yourself as you pick yourself up off the floor …] The Guardian has a story on how the Home Office used ‘AI hallucinated’ information to refuse an asylum claim. Twice. Essentially the refusal rested on a report denying the possibility of persecution on return. But that report did not exist. It just defies belief. Incredibly fortunate that the claimant managed a second challenge. But it shouldn’t depend on luck. And there really should be consequences for using AI like this. It’s outrageous. There should also be consequences for people submitting AI-based papers to academic journals. But that’s a separate story.
There are also a series of stories regarding the Met police’s plans to deploy a city wide drone network over London, the potential installation of facial recognition via lamppost cameras, and increasing use of facial recognition in shops. I do believe that the surveillance infrastructure for the application of advanced analytics is being put in place, in terms of pattern of life analysis, profiling, geo-fencing etc. Without any justification or, indeed, explicit legal basis. Not great.
In national international law news, the UK Supreme Court handed down its decision in The Kingdom of Bahrain (Appellant) v Shehabie and another (Respondents) this week. Its a really interesting decision which revolves around State immunity, and whether the use of spyware by a foreign State operating abroad but targeting an individual within the UK can give rise to an exception to that immunity. Tl’dr it can.
On the IHL side of things Airwars and AI Now have produced ‘Anatomy of a Kill Chain’ which is a visual representation of a fictionalised AI targeting cycle.
And (very worryingly) there is also the story of the OpenAI tool that went rogue, bust out of its ‘secure’ test environment, and hacked another company. Quite a few stories on this below.
This week i’ll leave you with Late Night Tales: Barry Can’t Swim. I’ve really been enjoying this mix over the last while. Very chill. Honourable mention to the Flaming Lips, and ‘Five Stop Mother Superior Rain’. Mainly for the ‘somebody please tell this machine i’m not a machine’ line – on brand for the newsletter.
Be well. Stay lovely.
— —
UK Supreme Court, The Kingdom of Bahrain (Appellant) v Shehabie and another (Respondents),
The Guardian, Home Office used ‘AI hallucinated’ information to refuse asylum claim, judge suggests oops
Airwars/AINowAnatomy Of An AI Kill Chain
TechSpot, US prosecutors charge Atlanta man after GrapheneOS phone wipes itself during airport search
The Guardian, Government use of automated AI decision-making to be curbed under new Australian rules
TechPolicy Press, The Silent World Cup Winners are the Companies Normalizing Surveillance
London Policing Ethics Panel Report, AI in the Metropolitan Police Service, July 2026
Lawfare, How the U.S. Military Learned to Embrace AI Warfare
The Guardian, ‘Adversarial clothing’: are garments designed to confuse facial recognition systems about to go mainstream?
The Guardian, Spies in the sky: how worried should we be about the arrival of AI-enabled smart lamp-posts?
EFF, An Explosion of Surveillance Towers is Coming to U.S. Borders, Costing Over $1 Billion
Open access ai articles, Cambridge Journals, AI and the Law,
ASPI, An AI busted out and ran amok. We should be scared
The Hill, OpenAI says its technology hacked another company in ‘unprecedented’ event
The Hill, New York school pauses plan to deploy humanlike AI robot teacher after backlash
BBC, Drone network part of Met Police’s London-wide tech push
Montgomery Advisor, Alabama family sues OpenAI, claims ChatGPT influenced woman’s death
Algorithm Watch, Click, Strip, Repeat: Sex Workers and Digital Violence Amidst the Deepfake Boom
BBC, Facial recognition vans to be deployed by Nottinghamshire Police
BBC, One arrest after police scan 50,000 faces at Appleby Horse Fair (one arrest, for breach of bail)
WIRED, The ACLU Is Arming Lawyers to Expose State Surveillance Secrets
WIRED, OpenAI’s Hacking Debacle Was a Human Mistake
WIRED, OpenAI’s Rogue AI Agent Hacked More Than Just Hugging Face
The Conversation, How an OpenAI safety test became a real-world cyberattack on the Hugging Face platform
The Guardian, Rogue OpenAI agent that hacked startup tried to attack other firms
New York Times, What Is Open-Weight A.I.?
New York Times, Microsoft Unveils A.I. Cybersecurity Tools
The Conversation, Al-Qaida and the Islamic State are both adopting AI – but differ in how they think about the technology
The Washington Post, Cheaper, open and intelligent: Chinese AI models gain ground, as they make inroads in the US
The Financial Times, Universities face difficult choices over how to integrate AI