More in today's MTUK: pro.politico.eu/news/200873

On age verification, ORG is concerned about a race to the bottom as third-party vendors offer cheap services at the expense of robust security and accuracy, and the framing of privacy-enhancing technologies as "circumvention" tools.

Report also argues that ambiguity over what precisely constitutes acceptable content encourages platforms to "err on the side of disproportionate caution" by preemptively over-moderating perfectly legal expression, with no clear means of redress for those affected (the so-called "bypass strategy")

eg

+ risks to users (like skills fade),
+ unsafe use (eg under-trust of AI outputs, something that can be seen in the evaluation),
+ & new failure modes (hey, could Scottish govt end up relying on AI tools built by UK govt to do policy work? can't see how that could go wrong...)

Transparency could also be a double-edged sword here: @gavinfreeguard.bsky.social says openness needs to be balanced against risk of opportunistic stakeholders figuring out ways to game the system.

The subjunctive is doing some heavy lifting there (that much-queried £45bn in public sector productivity savings stat is still doing the rounds) and DSIT didn't use a control group, looking instead at past data on how long it took officials to thematically sort consultation responses.

Hi Tony, I'm a (trans) reporter at Semafor looking into this. Wondering if you're willing to chat.

Heya, am a (trans) reporter at Semafor looking into this. Wondered if you might be willing to chat.