AI Score raises $5.4M to keep enterprise AI agents in check
AI Score, a London startup, has raised $5.4M — about £4M — in seed funding for software that does something increasingly urgent inside big companies: keep track of where AI is actually being used, and who is responsible for it. The round was led by Fuel Ventures, with the founding investor GALLOS Technologies and a group of angels taking part, and follows a roughly $1M pre-seed the company raised out of stealth in late 2025.
The problem AI Score is chasing is the gap between an organisation’s AI policy and its AI reality. Employees, systems, and now autonomous AI agents are adopting generative tools faster than most companies can map, let alone govern. AI Score’s platform sits in that gap: it continuously discovers where AI is being used across the business, assigns each use an owner and a risk profile, monitors what those systems and agents are doing, and keeps an audit trail — in effect a live register and guardrail layer for a company’s whole AI estate. As agents move from demos into real workflows, knowing what they are touching stops being a nice-to-have.
The founding team leans hard on national-security pedigree. Chief executive Alex Harland was on the founding team of the National Cyber Security Centre, and co-founder Benita Tibb is a former City lawyer; the advisory bench includes the former GCHQ director Sir Jeremy Fleming, Starling Bank chair Colin Bell, and Darktrace co-founder Nick Trim. It is a lineage that keeps recurring in British security-and-governance AI — the same intelligence-community heritage behind companies like Prevalent AI — and it is a genuine edge when the buyers are risk and security leaders.
It also sits neatly against a story we covered last month. When the AI Security Institute found agents going off-script in its tests, it underlined why enterprises are nervous about turning agents loose. Governance tooling like this is the commercial answer to that nervousness: not slowing AI adoption down, but making it legible enough that a company will actually sign it off. It is an unglamorous layer, but a necessary one — and, increasingly, a category of its own.
Read the original story on Tech.eu .
Companies in this story
Previous coverage
- Prevalent AI raises $22M — nine years in, and profitable 30 August 2026
- UK AI Security Institute finds agents going off-script in cyber tests 13 August 2026
- A month-old DeepMind spinout is reportedly raising $700M for AI world models 24 September 2026
- Magentic raises £13M to put AI 'digital workers' into factory procurement 22 September 2026
- Resolutiion raises £8M to catch commercial disputes before they blow up 22 September 2026