The Fallen by Louise Brangan review – an enraging account of Ireland’s Magdalene laundries The horrifying story of the Catholic-run institutions that incarcerated thousands of women and girls

www.theguardian.com/books/2026/a... Louise Brangan's upcoming book The Fallen is absolutely essential - among the best works of nonfiction I have ever read.

The Much Maligned British State, not the Melton Mowbray Building Society, that is

And as @ciaranm.bsky.social says (and as someone who myself also worked for the MMBS for a number of years) it’s always v nice when the public sector puts out high quality analytical work like this!

Really interesting piece of work - and conclusions broadly the same as our new exploratory study on adoption of models in the cybercrime underground (don’t panic - yet) arxiv.org/pdf/2603.29545

Our evaluation of Claude Mythos Preview’s cyber capabilities | AISI Work We conducted cyber evaluations of Anthropic’s Claude Mythos Preview and found continued improvement in capture-the-flag (CTF) challenges and significant improvement on multi-step cyber-attack simulati...

This paper represents a small but deeply impressive and genuinely important achievement by the much maligned British state in what is probably the most important global issue of our era.

Hear me out ( 🧵) 1/

www.aisi.gov.uk/blog/our-eva...

This isn’t the occasion to debate its contents.

But the TL:DR summary is that Mythos is really good at hacking but it’s a case of prepare, don’t panic.

that’s a really useful dose of realism (not least because of the argument that it served anthropic’s commercial interests to hype up mythos) 11/

Issues 41 - 60 of Russian Hacker or "Khaker: Computer Hooligan Magazine" (‘‘XAKEP’’) : XAKEP : Free Download, Borrow, and Streaming : Internet Archive Issues 21 through 40 of Russian hacker magazine ‘‘XAKEP’’, from 200 through to 2002. A publication for hackers, or “khakeri”, made by...

#History of the #hacking scene in 2 countries is a mini fixation of mine, #Russia & #China. In an effort to make this history more accessible I have been uploading archives of #hacker zines from both countries to Internet Archive, as part of this, issues 41 - 60 of XAKEP.
archive.org/details/XAKE...

Screencap from Ghost in the Shell Stand-Alone Complex showing Batou and a Tachikoma. The shot has been altered to show the Tachikoma handing Batou a copy of the preprint.

Sneak preview of a new preprint looking at adoption of Generative AI in the cybercrime underground: arxiv.org/pdf/2603.29545 we use a very large crime forum dataset to explore how cybercrime actors are adopting (or not) these tools - a phenomenon we call 'vibercrime'

@addiemcgowan.bsky.social has done (and is doing) extremely cool work on the material political economy of the ad networks that these activities are all clustering around - e.g. www.tandfonline.com/doi/pdf/10.1...

Real chance that this going to impose big additional costs on platforms. Version of this where it starts to undermine the platform economy/ad model (which always had a cat-and-mouse with deviant SEO). V hard to police - there is a big trade on the crime forums in e.g. OpenAI, Anthropic accounts etc.

Thing we really saw utility w/ (until looming AI financial crash comes) was automating crap SEO, fraud, crypto scams etc. Striking reading reports of Silicon Valley coders using agents for 'side hustles' - often the same 'passive income'/dropshipping/MLM stuff we see in the cybercrime ecosystem.

But our empirical data shows v similar scalar contradictions in the cybercrime underground as in mainstream start-up ecosystem - 'pilots' failing to scale, apparently useful stuff saturating quickly, misalignment in underlying economics (see recent @edzitron.com and others).

And in terms of contributing to contradictions emerging in the material political economy of Gen AI - we do see them already report some success at manipulating the outputs of the big models. As they already have armies of bots, SEO pages, reddit accounts, gives them a direct line to training data.

This piece also contains one of my favourite quotes we've had in a paper recently - "If I was employed as a threat analyst and this is... behind next gen malware I'd be celebrating." (rest is in image). Real disdain in a lot of the cybercrime underground for GenAI tools as coding assistants.

Gone to Waste: The 'CVE' Industry After 9/11 After the rise of the Islamic State, I was part of this race for cash, working on CVE projects from media messaging to radicalization research, and it slowly dawned on me how flawed the industry is. N...

@lsmwilson.bsky.social has written (and is writing!) a ton of brilliant stuff on this - eg newlinesmag.com/argument/und...

So you could train a bunch of extremism/derad specific models - but even worse risks there. Extensive and well documented history of blowback, looping effects, which we largely still don’t actually know how to systematically predict or prevent, let alone the big implications for rights, democracy

Influence Policing: Domestic Digital Influence Campaigns and Algorithmic Strategic Communications in UK Law Enforcement and Homeland Security Abstract. This paper conceptualizes an emerging model of algorithmic policing; ‘influence policing’. This harnesses the affordances of Internet platforms t

academic.oup.com/bjc/article/... We wrote about (and in some cases evaluated) the earlier generation of this stuff - v specialised targeting across lots of areas. Chatbots suggest more individualisation but obv colossal risks to users as these conversations will be well outside central distribution

More Redirect method stuff moving to chatbots. Influence Policing (domestic influence interventions carried out by police and security services for crime and terrorism prevention) largely still in the domain of behaviour targeted ads but future seems split between chatbots and $$ human influencers.

Thanks Lee! Interesting patterns of usefulness too (not much utility for hacking at that level but some real improvements in logistics, esp as the communities operating from a fairly crap material base). And v good for human-ifying patterns from automated processes and so passing defender detection.

But also - the big use cases they are finding are in logistics and in masking patterns from existing forms of automation, not really exploit development/hacking

FIPR very pleased to sign this statement by EDRi against current moves by the US govt to rein in EU regulation of technology providers edri.org/wp-content/u...

VIBERCRIME

To bastardise a quote:

“One should be aware of the impact that one has, in all ways, and not just those ways that one intends”

Important point here for the future of #Cybercrime and #DigitalPolicing
#Criminology

A lot of (often very good) studies to date showing massive gains in lab simulations - but there is a big difference in how they translate to real business models, saturation, and the reality of low skilled 16 year olds’ actual ability to use them to any effect.

But - we do see a big surge in people on the cybercrime fora saying that they have just lost their day job in IT because of AI layoffs. So as these more skilled IT workers look for other (including illicit) sources of income, that may itself have a major effect on cybercrime patterns!

I should also add that in the absence of a CEO or VC class in the cybercrime underground to push hype, most of the chat at least up to Dec 2025 is that LLMs and chatbots and agents not really that useful. So even the script kiddies are largely not buying it.

Ah thanks James! Right back at you :)

Ben Collier on Tor: From the Dark Web to the Future of Privacy Podcast Episode · Peoples & Things · March 30 · 1h 1m

Great interview with my brilliant colleague @susansegfault.bsky.social about his excellent book on Tor-Dark Web. highly recommend both the interview and the book, which is available free at MIT press direct.mit.edu/books/oa-mon...

Screencap from Ghost in the Shell Stand-alone Complex depicting three Tachikomas and the Jameson robot. Subtitle reads "Fights and pranks, bring them on!"

We'll do a proper release of the paper soon (it's been peer reviewed and we're presenting it later in the year) - it's been great working with Jack Hughes (Cambridge) and Daniel Thomas (Strathclyde) on this as ever, and w/ thanks and love to Prof Alice Hutchings and the Cambridge Cybercrime Centre.

Screencap from Ghost in the Shell Stand-alone Complex depicting a senior police officer at a press conference, with the Laughing Man logo superimposed over his face. He is making a gun with his fingers.

Screencap from Ghost in the Shell Stand-Alone Complex showing the Jameson robot

We argue that genAI is largely being adopted for criminal logistics, not for hacking. Two possible trajectories - the Stand-Alone Complex (genuine scale transition from automation) and vibercrime (adoption for vibe coding further proletarianises the work of cybercrime-as-a-service).