The deeper resources behind the briefings. Every asset is short, sourced and free. Filter by type, or by the three themes that matter most when applying AI at scale: value, risk and trust.
New guidance on GOV.UK gives public sector teams a common way to identify, score and treat AI risks across nine categories. It is not mandatory, but it sets a clear benchmark.
The FCA’s review of how firms are using the most advanced AI models for cyber security concludes that the hardest part is not the technology but organisational readiness.
The National Cyber Security Centre has set out interim advice on managing the cyber risk of AI agents. Its core message: assume an agent will sometimes act in ways you did not intend, and design for that.
How a statistically defensible model lost public legitimacy in four days
What happens when an algorithm becomes the business model faster than anyone can learn from it
A rare example of clinical AI that made it into daily use, and why the people around it mattered as much as the model
Why removing the obvious signals of bias was not enough to remove the bias
What happens when generative AI is used to publish at volume, quietly, with human review that does not catch the errors
How a flagship AI promise met the reality of clinical evidence, hospital systems and expert judgement
How Google moved from AI recommendations to AI control of a critical system, with safety designed in from the start
When a system can be accurate at its narrow task and still fail to deliver the outcome it was bought for
Impressive efficiency figures in month one, and a change of course fifteen months later
A small claim with a large lesson: an organisation is responsible for what its AI tells customers
What a government department’s own fairness test found, and what it had not yet tested
How fraud risk profiling, rigid enforcement and nationality data combined into an “unprecedented injustice”
World-class research results, a hard road to routine care, and a lesson in data trust from another NHS trust
Map your AI services against ownership, control, jurisdiction, capability and optionality, and see which need attention first.
One short email when we publish a new briefing, case file or tool. Free, and easy to unsubscribe.
Infotactica