Aligne.AI

We help enterprises govern AI in production, not just on paper. AI Governance Platform + practitioner expertise serving enterprises across the UK, EU and Middle

[Aligne.AI @Aligne_AI]

Full data and what it means for building a programme now: https://t.co/xn7O74g0z6

#AIGovernance #UKTech

[Aligne.AI @Aligne_AI]

44% of large UK businesses already run AI in production.

17% of boards say they own AI oversight at all.

7% call their governance fully embedded.

That gap is not unique to finance.

[Aligne.AI @Aligne_AI]

76% of UK teachers use AI. 76% report zero training on it. 49% of schools have no AI policy at all. (Teacher Tapp/Jisc)

Every sector, same pattern: adoption outrunning governance.

[Aligne.AI @Aligne_AI]

More on what surprised us going through all 99: https://t.co/COMHQ0pxRJ

#AIGovernance

[Aligne.AI @Aligne_AI]

IBM's AI Risk Atlas names 99 specific AI risks across 5 categories. Studying it convinced us a generic risk checklist is nearly useless.

"Do you have privacy risks" gets a yes from everyone. That tells you nothing.

[Aligne.AI @Aligne_AI]

"Confidential data in prompt." "Sharing IP with tools." Those get a specific, checkable answer.

We are building our own AI Risk Library on that principle: named risks, named regulations, named controls, not a rebadged checklist.

[Aligne.AI @Aligne_AI]

"We reviewed the model" used to be enough. US regulators just excluded agentic AI from model risk oversight, calling it "novel and rapidly evolving." Gartner: 40%+ of agentic projects scrapped by 2027, mostly governance failure, not cost.

https://t.co/zR0kUJAURV

#AIGovernance

[Aligne.AI @Aligne_AI]

An AI risk register is not a compliance PDF filed once and forgotten. It is the one document examiners ask for first: every AI system in use, its risk level, the controls attached to it.

Most orgs either don't have one, or built it once and never opened it again.

#AIGovernance

[Aligne.AI @Aligne_AI]

More on what surprised us going through all 99: https://t.co/COMHQ0pxRJ

#AIGovernance

[Aligne.AI @Aligne_AI]

7% of large UK firms call their AI governance fully embedded. 17% of boards say they own AI oversight at all.

If your board asked today, what is the honest answer?

#AIGovernance

[Aligne.AI @Aligne_AI]

Not every AI system is “high-risk” under the EU AI Act. More of yours probably qualify than you think.

Category membership is only half the test. One exception-canceling catch: if the system profiles individuals, it is high-risk regardless.

https://t.co/vPwFWblr35

[Aligne.AI @Aligne_AI]

Runtime: everything monitored after. Continuous, not one-time.

The gap between them is where most real risk actually shows up.

[Aligne.AI @Aligne_AI]

Design-time and runtime get used constantly in AI governance conversations, almost never defined.

[Aligne.AI @Aligne_AI]

Design-time: everything reviewed before launch. A snapshot, taken once.

[Aligne.AI @Aligne_AI]

The EU AI Act’s biggest deadline just moved again.
Annex III high-risk rules, due 2 August, got pushed to December 2027 six days before deadline.

Transparency rules did not move. Live since 2 August.
One moving did not mean the other did.

#EUAIAct

[Aligne.AI @Aligne_AI]

Three things a model card cannot show you: what it actually did, whether it has drifted, what happens when it chains actions together.

[Aligne.AI @Aligne_AI]

A model card tells you what a system was trained on and how it performed on a fixed set of tests.

[Aligne.AI @Aligne_AI]

None of that tells you what an AI agent is doing right now, in production, with inputs nobody tested for.

[Aligne.AI @Aligne_AI]

A model card tells you what a system was trained to do.
It cannot tell you what an agent actually did today, whether its behaviour has drifted, or what happens when it chains actions nobody reviewed together.

https://t.co/kVyVNff4Ob