AI Policy Update: 22 July 2025
22 July 2025
Three things this week
First, AI scheming, gold-standard research, and the impact on policy.
The UK AI Security Institute has released a new research paper: "Lessons from a Chimp: AI 'Scheming' and the Quest for Ape Language". The paper examines research that asks whether current AI systems may be developing a capacity for covertly and strategically pursuing misaligned goals, or 'scheming'.
As Keegan McBride of the Tony Blair Institute for Global Change notes, the paper argues that emerging AI research often relies on claims that hinge on anecdotal evidence and maybe interpreted in exaggerated ways. So "gold standard science" is critical.
As policymakers consider the societal impact of the technology, AI research must remain objective.
See Keegan's excellent post for more: https://lnkd.in/emwWABwA.
Next, AI, automation and political realignment.
In an excellent and thought-provoking Substack, James Kanagasooriam, Chief Research Officer at Focaldata, discusses the rise of AI and its potential impact on UK politics. (h/t to Jimmy McLoughlin OBE for the initial post).
In the Substack, James says that "Executive Britain" appears to be in for an incredibly tough decade. Based on his modelling, the top constituencies with the highest job exposure to AI and automation include Richmond Park, Highgate & Hampstead, and the Cities of London and Westminster.
Compounded with sky-high house prices, student debt and high interest rates, James suggests that "automation may be the final straw...before white collar Britain goes into full revolt."
Given the potential scale of impact, it's a must-read piece, especially for those considering the policy response to the technology.
Last, the EU General Purpose Code of Practice.
Narrowing missing last week's update, the EU published the final version of the code for general-purpose AI models on 10 July.
Here's a selection of what experts on this platform are saying about it:
Julia Mykhailiuk from Pour Demain: https://lnkd.in/e6FnHnfZ
Jonas Schuett from Centre for the Governance of AI (GovAI): https://lnkd.in/eM2e6w4J
Risto Uuk from the Future of Life Institute (FLI): https://lnkd.in/eymmznEU