AI Score Raises $5.4 Million as Security Alumni Watch Agents

AI Score has raised $5.4 million (£4 million) in seed funding to scale a live control layer for generative and agentic AI inside companies.

Fuel Ventures led the round, founding investor GALLOS Technologies returned, and the deal was recorded on 3 September 2026. The cheque is security and City capital buying a watchtower, because the people who used to run Britain’s cyber defences think ordinary governance cannot see what agents do.

GCHQ Alumni Now Watch Corporate AI Agents

Alex Harland is chief executive. He was part of the founding team at the UK’s National Cyber Security Centre, later served as a National Incident Manager there, and ran cybersecurity at Compare The Market and BGL Insurance. Benita Tibb is chief operating officer, a former City lawyer who was an associate at White & Case and chief of staff at GALLOS.

The National Cyber Security Centre formed in 2016 as GCHQ’s public-facing cyber defence arm. Harland’s path from that shop to a Paul Street software firm is the product thesis: agents now act across tools, people and outside systems, and a quarterly policy pack cannot keep up.

Sir Jeremy Fleming, Director of GCHQ from April 2017 to May 2023, advises the company. So do Colin Bell, chair of Starling Bank and former head of HSBC Bank Europe; Nick Trim, a Darktrace co-founder; Jonathan Kewley, partner and co-chair of the global tech group at Clifford Chance; and Josh Burch, GALLOS co-founder and a non-executive director at AI Score. That is a board-and-security circle, not a typical seed advisory slide.

AI operates at a scale and level of automation that traditional governance simply wasn’t built for. Organisations need to be able to move quickly and capture the opportunity AI presents, without losing visibility or control. This investment allows us to accelerate our ambition to reinvent governance for the AI age, building intelligence directly into the process and giving organisations the confidence to win with AI.

Alex Harland, co-founder and CEO, AI Score

Burch put the same worry in plainer terms when the firm came out of stealth. Large parts of the corporate world, he said, have not yet absorbed the risks they are running.

Fuel Ventures Led a 95th-Percentile Seed

The seed follows a $1 million pre-seed in November 2025 and takes disclosed funding to $6.4 million. Dealroom ranks the new round around the 95th percentile by size among UK enterprise software seed deals on record. PitchBook lists seven investors.

Fuel Ventures is a London firm whose founder page stresses a vertical AI and AI-enabled software brief. GALLOS came back. So did named angels from finance and security: Alan Morgan, co-founder of MMC Ventures and an early backer of Funding Circle and Tide; Mo El Husseiny, managing partner of Ventura Capital; and Robert Mann, founding partner at Marshall Bridge Ventures.

THE SEED ROUND’S BACKERS

The company said revenue rose sharply in the first half of 2026, driven by regulated and data-heavy clients, including a law firm, global consumer brands, a leading UK fintech, FTSE 250 companies and Magic Circle firms. It declined to name most of them in the round notes.

What the Platform Watches

AI Score sells an intelligence-led governance layer meant to live between written policy and operational reality. The firm calls the space in the middle the AI Governance Reality Gap: unknown risk, uncontrolled exposure, missed value and governance you cannot prove. On one side sit policies, controls, objectives and risk appetite. On the other sit people, platforms, agents and workflows that change every day.

The product reads unstructured evidence (prompts, outputs, context, actions), scores it against the firm’s own objectives, and turns that into ranked actions. It is private-cloud SaaS, API-driven, and pitched as live in minutes rather than a months-long GRC rollout. For agents, the pitch is specific: watch them as they arrive, set what they can do and connect to, and keep a trail management and boards can actually see.

THE SIX GOVERNANCE LAYERS

• Live policy: Connect paper rules and objectives to real AI activity and show where they break.
• Safe adoption: Replace blanket blocks with precise guardrails so staff can use AI without a free-for-all.
• Asset register: Give every model, tool and agent an owner, a risk profile and a live audit trail.
• Human usage: Track intent and behaviour so literacy, cost and adoption can be measured.
• Agent control: Limit what agents can do, access and connect to, and flag drift as it happens.
• Board picture: Fold risk, cost, intent, usage and objectives into one view of where to act.

Agent work already splits into two markets: the agents themselves, and the systems that can say what those agents did. A policy PDF does not cover the second job. Buyers who have tried to run agents want controls that travel with the work, with live task state, scoped permissions and a trail you can replay, not a binder that is updated after the fact.

Clifford Chance Put the Guardrails on Record

The company site names Clifford Chance as a user, which is rare in a seed story that otherwise keeps customers off the record. Anthony Vigneron, Director of Legal Technology Solutions at the firm, is quoted there on guardrails and a continuous view of every AI interaction.

AI Score empowers us to establish and maintain essential guardrails on our AI usage across our global operations. It provides unprecedented visibility into where and how AI is utilised throughout our firm. By continuously monitoring and evaluating every AI interaction, AI Score ensures the technology’s safe, compliant, and effective deployment.

Anthony Vigneron, Director of Legal Technology Solutions, Clifford Chance, on AI Score’s site

Kewley, a Clifford Chance tech partner, also sits in the startup’s advisory group. Mark Hendry, a partner at S&W, is quoted on the same page saying the platform lets the firm show clients their AI usage and risk faster than rival advisers. The early beachhead is law, consumer brands and UK fintech, places where a leaked prompt or a rogue agent is a commercial event, not a lab demo.

Colin Bell said the company is building an approach that can turn governance from a constraint into a driver of value. That line only works if the software is fast enough that security teams stop being the department of no. The seed is a bet that they will pay for that speed.

Three in Four Firms Plan Agents Without Mature Controls

Deloitte’s State of AI in the Enterprise work is the cleanest public measure of the gap AI Score is selling into. A 12 August 2026 press note on that survey said 74 percent plan to deploy agentic AI within two years, while only 21 percent report a mature model for governing autonomous agents.

THE AGENT GOVERNANCE GAP

• Intent to deploy: 74 percent of companies in Deloitte’s survey plan to put agentic AI to work within two years.
• Mature controls: 21 percent say they already have a mature governance model for autonomous agents.
• Survey base: The wider State of AI in the Enterprise study polled more than 3,000 business and IT leaders across 24 countries.
• The stall risk: Deloitte’s note treats that gap as a live operational risk as firms scale systems that can act, not just answer.

Harland’s “traditional governance” line is the vendor version of that finding. Boards are signing off on agents that can call tools, write to systems and talk to people, then discovering their control set is still a register, a forum and a yearly audit. The people writing cheques into AI Score are betting that gap stays open long enough to build a category, and that regulated buyers will pay to close it first.

A Defence Studio Built the Company Then Returned

AI Score did not start as a two-founder garage. It was built inside GALLOS Technologies, a studio set up in 2021 by Burch, a former senior UK national security official, and Dean Jones, an armed forces veteran. GALLOS backs and builds defence, security and resilience companies. In August 2026 it raised $50 million, in a round still described as open, led by Ventura Capital and Aberdeen Investments, with Lansdowne Partners participating. Fleming chairs GALLOS’s strategic advisory board. Burch’s own biography records more than twenty five years in intelligence and government, plus a later stint leading FTI Consulting’s cybersecurity practice in EMEIA.

When AI Score left stealth on 10 November 2025 with more than £750,000 of pre-seed capital (the round was also described as $1 million), Burch was already on the advisory board. He is now a non-executive director. That is the hidden stakeholder in this seed: a defence studio recycling national-security networks into enterprise software, then doubling down once the product had regulated logos.

FROM STUDIO BUILD TO SEED

1. 3 October 2016: The National Cyber Security Centre opens as the UK’s public cyber defence hub, the institution Harland later cites as his founding-team credential.
2. 2021: Burch and Jones set up GALLOS to back and build defence, security and resilience firms.
3. 2025: Harland and Tibb found AI Score inside that studio.
4. 10 November 2025: The company emerges from stealth with a $1 million pre-seed and unnamed design partners in law, consumer and fintech.
5. First half of 2026: Management says revenue rises sharply among regulated, data-sensitive clients.
6. August 2026: GALLOS raises $50 million for its next cohort and platform, with Fleming still in the chair.
7. 3 September 2026: Fuel Ventures leads a $5.4 million seed; GALLOS returns.

Fleming, speaking around the GALLOS raise rather than the AI Score seed, said the world faces growing digital threats sharpened by the speed at which new technology is turned against people and businesses. That is the same clock the startup is selling to general counsel.

The Cheque Pays for Agent Controls and a Sales Push

Management said the seed will accelerate platform development and expand go-to-market capacity, with a particular focus on tools for managing agentic AI as it spreads through the enterprise. That is a product choice as much as a budget line. Generative chat is already inside most large firms. Agents that can act are the new sprawl, and they are what a former incident manager is wired to worry about.

The round does not disclose a valuation, a headcount target or named new hires. PitchBook still files the company under network management software, which is a dry way of saying it wants to be the control plane, not another chatbot. Fuel led. GALLOS stayed. Fleming’s name is on the door. Clifford Chance is already on the record. The rest of the work is to make boards pay for a live view of agents they have already let in.

Frequently Asked Questions

Where is AI Score based?

PitchBook lists the office at 86-90 Paul Street, 3rd Floor, London EC2A 4NE, and classifies the company as network management software. The firm was founded in 2025 and remains privately held.

How is AI Score’s software deployed?

The company describes a private-cloud SaaS product that is API-driven and designed to be live in minutes rather than a long GRC implementation. Staff can type a policy in plain English, and the platform is meant to turn that text into enforceable controls across the AI estate.

Did AI Score disclose a valuation for the seed round?

No valuation was published with the $5.4 million seed. Dealroom’s ranking is about cheque size versus other UK enterprise software seeds, not about the price of the shares.

What size of seed cheque does Fuel Ventures usually write?

Fuel’s founder page lists seed cheques of £1 million to £3 million for revenue-generating businesses, about 10 to 15 deals a year, across a portfolio it puts at 230-plus companies. AI Score’s headline £4 million round is larger than that solo band because several other backers joined the lead.

Who were AI Score’s first design partners?

When the company left stealth in November 2025 it said it was already working with a magic-circle law firm, a major consumer brand and a fast-growing AI-enabled fintech, without naming them. Clifford Chance has since gone on the record on the company site; the other logos from that launch list have not been published.

The new capital is earmarked for agent controls and a larger sales bench. Until boards can see those agents in the same window as their policies, the watchtower still has work.