UK organisations are embedding AI governance into daily operations faster than their global peers, yet most remain unable to reconstruct and defend AI-assisted decisions when regulators, auditors or legal teams demand answers, according to new research from Arctera, a Cloud Software Group business unit that supplies compliance and governance solutions worldwide.
The findings appear in State of AI Governance 2026: The UK Perspective – Where Operational Governance Meets the Reconstruction Challenge. The report shows British firms outpacing global averages across several core controls.
Human sign-off on high-risk AI communications is mandatory at 89% of UK organisations, against 79% worldwide. Training on compliant AI use reaches employees at 85% of UK firms, compared with 77% globally. Formal AI usage policies exist at 83% of UK organisations versus 80% elsewhere, while 77% track or record AI prompts and outputs, well above the 65% global figure. Routine audits of AI tools and outputs take place at 70% of UK firms, compared with 63% globally.
Arctera cautions that deploying controls does not settle the matter. Once an audit, investigation or legal dispute arises, firms must be able to demonstrate with clear evidence that those safeguards actually functioned.
That proves difficult because the records generated by AI controls are typically scattered across separate systems, departments and communication channels. Four in ten UK respondents named linking communications across formats and channels as a significant obstacle.
Exposure rises sharply when staff turn to consumer messaging apps or personal calls. Some 73% of UK organisations describe their surveillance of these channels as partial or absent, with the figure standing at 59% for personal phone calls and 27% for text messages.
The consequences are tangible. Eight in ten UK organisations, 80%, cannot open an investigation with complete information to hand, and a third devote over half of their investigation time to collecting and reconciling data before any analysis starts.
Arctera labels this the “reconstruction challenge”. When a decision, communication or record comes under scrutiny, teams often must manually stitch together the AI activity, reviews, approvals, edits and conversations that produced the final result, an administratively heavy exercise.
The UK analysis is drawn from Arctera’s broader State of AI Governance 2026: The Policy-to-Proof Gap in AI-Assisted Communications, an independent quantitative benchmark that Hanover Research carried out in May 2026 using a pre-vetted industry panel.
It covers 100 UK participants from the 500-person global survey, all full-time compliance decision-makers or influencers in finance, healthcare and energy or utilities, working at organisations with at least 500 staff and $50m in yearly revenue. Results are unweighted and directional, and no traditional margin of error applies given the nonprobability sample.
Arctera SVP and general manager Soniya Bopache said, “UK organisations are moving beyond simply setting rules for AI use. They are increasingly putting governance into practice through employee training, human review, prompt and output logging, retention and auditing.
But controls alone are not enough. When scrutiny comes knocking, organisations need to be able to connect the evidence and show quickly and confidently what happened from beginning to end. This is central to defensible AI governance.”
Copyright © 2026 RegTech Analyst
Copyright © 2026 RegTech Analyst
