22one
Building AI tools for police and law-enforcement work, operating in stealth mode.
Website: https://22one.ai/
Cover Block
Public sources
| Field | Value |
|---|---|
| Name | 22one |
| Tagline | Building AI tools for police and law-enforcement work, operating in stealth mode [Dealroom News, October 2026] [22one] |
| Headquarters | London, UK [Inc., October 2026] |
| Founded | 2025 [Dealroom News, October 2026] |
| Stage | Seed [Funding Spotter, 2026] |
| Business Model | B2B |
| Industry | Defense / Govtech |
| Technology | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Solo Founder |
| Founder | Steven Richardson [Dealroom News, October 2026] |
| Funding Label | Undisclosed [Financial Times] |
| Total Disclosed | $14,000,000 [Funding Spotter, 2026] |
Links
Public sources
- Website: https://22one.ai/
Executive Summary
PUBLIC 22one is a London-based stealth startup building AI tools for police and law-enforcement work, and it merits attention now because public reporting says it reached an approximately $400 million valuation in October 2026 before publicly launching a product or naming customers [Inc., October 2026] [Financial Times] [Dealroom News, October 2026]. The founding story is unusually specific to the buyer: the company was founded in 2025 by Steven Richardson, a former senior UK police officer who later held public-safety roles at AWS and Palantir, giving 22one a clearer domain wedge than most early govtech AI companies can claim at seed stage [Reuters, February 2008] [RocketReach, Retrieved 2026] [Liverpool Echo, Retrieved 2026].
What the company appears to be building is still thinly documented. Public descriptions consistently frame the product as AI software for police and investigative workflows, while the company website itself remains minimal and presents only a stealth positioning rather than a product explanation [22one] [Inc., October 2026] [The Hustle, November 2026]. That leaves the main differentiator, for now, less in disclosed technology than in founder credibility, sector access, and prior exposure to public-safety software problems, including Richardson's earlier work with Riven and DocDefender [Dealroom News, October 2026] [LinkedIn, Retrieved 2026].
On capital formation, the picture is directionally strong but not fully reconciled. Financial Times-linked reporting says 22one attracted backing from Sequoia Capital, General Catalyst, Lux Capital, and Creandum at a roughly $400 million valuation, while Funding Spotter lists a $14 million seed round in 2026 and £15.9 million raised during the year, figures that should be treated cautiously until matched to primary filing or investor confirmation [Financial Times] [EUVC, October 2026] [Funding Spotter, 2026]. The business model is straightforward on paper, B2B software sold into police and law-enforcement agencies, but the harder underwriting question is whether 22one can translate founder-led access into procurement-grade deployments inside a slow, compliance-heavy public sector market [Dealroom News, October 2026] [Funding Spotter, 2026].
Over the next 12 to 18 months, the signals that matter are not incremental press mentions but evidence of execution: a clearer product surface, named deployments, proof that the software fits real investigative workflows, and any sign that the company can navigate public-sector adoption without remaining entirely founder-dependent [Inc., October 2026] [The Hustle, November 2026]. The investor interest is understandable given the team and valuation signal, but at this stage the public record still supports more confidence in the quality of the cap table than in operating traction.
Lightly corroborated -- This section relies on one widely cited Financial Times report, corroborating secondary coverage, the company website, and public-profile records; product detail and funding specifics remain only partially verified.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | B2B |
| Industry / Vertical | Defense / Govtech |
| Technology Type | AI / Machine Learning |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Solo Founder |
| Funding | Undisclosed |
How the Company Got Here
Public sources
22one appears to have been set up with an unusually thin public footprint for a company that surfaced quickly into late-stage valuation territory. Its own website offers only the company name and the phrase "In Stealth," which is consistent with the limited public record and the absence of a disclosed product launch [22one].
The chronology that can be stated with some confidence is short but relevant. Public reporting cited in October 2026 described 22one as a stealth startup founded by former senior police officer Steven Richardson and backed by Sequoia Capital, General Catalyst, Lux Capital, and Creandum, with the financing amount undisclosed and the company reported at roughly a $400 million valuation [Dealroom News, October 2026] [Financial Times]. Separate reporting also places the company's incorporation in December 2025, which fits the sequence implied by the October 2026 financing reports, though the underlying company filing is not included in the source pack here [Dealroom News, October 2026].
On the available public evidence, the operating picture is still narrow: a London-headquartered legal entity in stealth, a named founder, and a reported financing event less than a year after incorporation [22one] [Dealroom News, October 2026] [Financial Times]. That is enough to establish existence and timing, but not yet enough to map product maturity, customer rollout, or internal scale from public documents alone [22one] [Financial Times].
Lightly corroborated -- Confirmed in part by the company website and Dealroom coverage; some milestone details rely on a single reported financing narrative referenced by the Financial Times.
Product and Technology
MIXED
The public record on 22one's product is unusually thin, which is itself the main analytical fact here. The company website presents only the name, a meta description reading "In Stealth," and, according to newsletter reporting, a homepage that displayed the single word "investigate," with no published product documentation, feature list, pricing, or demo materials surfaced in the cited sources [22one] [The Hustle, November 2026]. Press coverage is directionally consistent in describing 22one as building AI tools for police and law-enforcement work, but those descriptions are still secondhand and do not establish what the software actually does inside an agency workflow [Inc., October 2026] [Dealroom News, October 2026] [Tech Funding News, October 2026].
The narrowest supportable read is that 22one is aiming at investigative and operational efficiency for police users, with the founder's prior work offering the clearest clue to product orientation. Multiple reports tie Steven Richardson to senior policing roles and later public-safety work at AWS and Palantir, while secondary coverage also links him to Riven's DocDefender, a tool used to automate redaction of sensitive personal information from police documents [Reuters, February 2008] [RocketReach] [Liverpool Echo] [LinkedIn, Retrieved 2026]. That background suggests a buyer-specific wedge around police data handling, intelligence, or investigative workflows, but any tighter statement about model architecture, deployment method, integrations, or production maturity would be inference rather than observation based on the evidence available [Financial Times] [Inc., October 2026].
Company-stated, unverified -- This section relies materially on company-minimal web presence and secondary reporting, with no verified public demo, technical documentation, or disclosed customer deployment.
Where the Demand Sits
PUBLIC
The market matters now because governments are moving from pilot-stage interest in public-sector AI to budgeted programs, while police and law-enforcement workflows remain document-heavy, labor-intensive, and politically exposed enough that procurement can accelerate quickly once a trusted use case is established [EUVC, October 2026].
Public evidence on 22one's addressable market is thin, so any market framing has to stay close to adjacent, named programs rather than pretend precision. The clearest public demand signal in the source set is the UK Home Office's PoliceAI initiative, which the available reporting says launched in June 2026 with 375 million over three years and planned pilots involving up to 10 police forces during 2026 to 2027 [EUVC, October 2026]. That is not a 22one contract, and it is not a market-size figure for police software as a whole, but it is a concrete, government-backed indicator that AI spending in policing has moved into funded experimentation.
A second demand indicator is the kind of founder-market fit investors appear to be underwriting. Multiple reports describe 22one as aiming to build AI tools for police and law-enforcement work, with Steven Richardson's prior policing, AWS, and Palantir background presented as a commercial wedge rather than a broad horizontal software thesis [Inc., October 2026] [Dealroom News, October 2026] [Tech Funding News, October 2026]. Read conservatively, that suggests investors see a category where domain access, procurement fluency, and trust may matter as much as model performance.
The adjacent markets are easier to define than 22one's exact product category. If the company is focused on investigation workflows, the nearest substitutes likely sit across records management, evidence handling, redaction, intelligence analysis, and case-preparation software, because those are the systems where police agencies already pay for productivity, compliance, and searchability [The Hustle, November 2026] [Dealroom News, October 2026]. Richardson's earlier work on automated redaction through Riven's DocDefender is relevant here, not as evidence of 22one's current product, but as a signal that privacy-preserving document workflows are one plausible spending wedge inside the broader law-enforcement software stack [LinkedIn] [Dealroom News, October 2026].
The regulatory picture cuts both ways. On one side, public-sector pressure to improve efficiency and investigator productivity supports adoption, particularly when agencies face staffing constraints and rising digital evidence volumes [The Hustle, November 2026]. On the other, policing is among the most scrutinized AI end markets, so procurement is likely to be shaped by auditability, evidentiary standards, privacy controls, and political acceptability, all of which can slow expansion beyond narrow initial use cases even when budgets exist [EUVC, October 2026] [Inc., October 2026].
| Cited market or adjacent-budget signal | Amount | Scope | Source |
|---|---|---|---|
| UK Home Office PoliceAI program | 375m over 3 years | Government initiative planning pilots involving up to 10 forces during 2026 to 2027 | [EUVC, October 2026] |
| Reported 22one funding in calendar 2026 | 315.9m | Company financing, included here as an investor appetite signal rather than market size | [Funding Spotter, 2026] |
| Reported 22one valuation in October 2026 | about $400m | Private-market pricing signal for police AI exposure, not end-market revenue | [Financial Times] |
The small set of public numbers says more about momentum than size. Government budget formation and investor pricing both point to a category that is opening, but neither establishes how large the near-term software spend available to a single vendor will be.
Lightly corroborated -- Section relies on one directly cited adjacent government-budget datapoint from EUVC and reported company financing signals from Financial Times and Funding Spotter; no independent third-party TAM study was surfaced in the provided public sources.
Competitive Landscape
Market map
MIXED The competitive set around 22one is defined less by direct startup peers, which have not been named in the available reporting, and more by the institutions and vendors that already sell into policing, intelligence, and public-safety workflows. Public reporting places 22one in stealth, building AI tools for police and law-enforcement agencies, with a minimal public web presence and no disclosed customer deployments or product detail [Inc., October 2026] [The Hustle, November 2026] [22one]. That means the practical alternatives today are likely to fall into three buckets: incumbent public-safety software providers embedded in records, dispatch, and case-management systems; analytics and decision-support platforms associated with government work; and internal police technology teams pursuing bespoke workflows under broader government AI programs, including the UK Home Office's PoliceAI initiative [Financial Times] [EUVC, October 2026].
The incumbent lane matters because police buying behavior tends to favor systems that already clear procurement, security, and audit requirements, even when product quality is uneven. The challenger lane matters because newer AI suppliers can move faster on redaction, search, summarization, and investigative workflow tools, especially where a founder has direct domain credibility. The adjacent substitute is not a startup at all in many cases, but a force choosing to extend existing vendors or use government-backed programs rather than adopt a new standalone product [Dealroom News, October 2026] [Inc., October 2026].
Edge and durability
MIXED 22one's clearest edge, on the public record, is founder-market fit rather than visible product proof. Steven Richardson previously served as head of the Met's Specialist Intelligence Section and was identified by Reuters in 2008 as a senior policing official, later holding policing-focused roles at AWS and Palantir according to RocketReach and Liverpool Echo [Reuters, February 2008] [RocketReach, Retrieved 2026] [Liverpool Echo, Retrieved 2026]. He also previously co-founded Riven, whose DocDefender product won a National Police Chiefs' Council Commissioner's Award with Bedfordshire Police, according to LinkedIn [LinkedIn, Retrieved 2026].
That background can matter more in this category than in ordinary enterprise software. Police customers are unusually sensitive to operational credibility, procurement trust, data handling, and the difference between a workflow that looks useful in a demo and one that survives evidentiary scrutiny. The edge is real if Richardson can convert prior relationships and category fluency into pilots before larger platforms narrow the gap. It is also perishable. Founder access and investor pedigree can accelerate first meetings, but without disclosed deployments, proprietary data assets, or a visible systems-of-record foothold, the moat still appears relational rather than structural [Financial Times] [Dealroom News, October 2026].
Exposure points
MIXED The company's biggest exposure is that the most credible alternatives may already control the channel 22one needs to win. Palantir is not confirmed as a direct competitor in the source set, but it is a named institution in Richardson's background and an obvious reference point because it already has government credibility, established procurement pathways, and a history of public-sector analytics deployments [Liverpool Echo, Retrieved 2026]. If a police force wants a broad intelligence and operational platform rather than a narrowly scoped investigative tool, a platform vendor with existing security and contracting infrastructure starts ahead.
22one is also exposed to substitution from internal modernization programs. The Home Office's PoliceAI initiative, cited in the research context, is not a disclosed 22one partner or customer, but it illustrates that some budget and experimentation may flow through centrally backed programs rather than to a standalone startup [EUVC, October 2026]. A stealth posture compounds that risk. While it preserves secrecy, it leaves the company without public proof points at a time when buyers, regulators, and the press are likely to ask for evidence on accuracy, bias controls, auditability, and operational fit before approving broader deployment [Inc., October 2026] [The Hustle, November 2026].
Eighteen-month scenario
MIXED The most plausible 18-month competitive scenario is a bifurcation between broad platforms and specialized workflow tools. Winner if procurement favors integrated, already-cleared public-sector platforms: Palantir, because its advantage would be existing institutional trust, implementation capacity, and a wider operating envelope across public safety and intelligence use cases [Liverpool Echo, Retrieved 2026]. Winner if police forces instead buy point solutions from operators with direct domain credibility: 22one, because Richardson's policing background and prior work around document redaction suggest a practical route into narrow, high-friction investigative tasks where usability matters as much as model performance [Reuters, February 2008] [LinkedIn, Retrieved 2026].
The loser if transparency demands rise faster than product maturity is the stealth entrant that cannot convert narrative into referenceable outcomes. On the public evidence, 22one carries that risk more than any named incumbent, simply because its valuation and investor roster are visible while its product, deployments, and customer evidence remain largely undisclosed [Financial Times] [Inc., October 2026] [Tech Funding News, October 2026]. That does not preclude a strong outcome, but it does mean the competitive case still rests on inferred access and founder credibility rather than demonstrated market control.
Lightly corroborated -- Competitive positioning relies on public reporting from Financial Times, Inc., Dealroom News, EUVC, Reuters, RocketReach, Liverpool Echo, LinkedIn, and the company website.
Opportunity
Public sources If 22one turns a founder-led credibility advantage into software adopted across police investigations, the prize is not a narrow point tool but a system-of-record layer for how law-enforcement agencies triage, investigate, and document casework across a high-trust, slow-moving buyer base [Financial Times, October 2026] [Inc., October 2026] [22one].
The headline opportunity is unusually clear even though the product is not. Public reporting consistently places 22one around a policing and law-enforcement workflow, with Steven Richardson positioned as a former senior police operator who later held public-safety roles at AWS and Palantir [Inc., October 2026] [Dealroom News, October 2026] [RocketReach]. That matters because police software is not a category where distribution usually begins with self-serve adoption. It begins with trust, procurement navigation, data sensitivity, and an ability to map software to operational reality. The reported investor roster, Sequoia Capital, General Catalyst, Lux Capital, and Creandum, does not prove product-market fit, but it does suggest sophisticated capital is underwriting the view that this founder can convert domain access into a large software company before the product is broadly visible [Financial Times, October 2026] [EUVC, October 2026].
A second reason the upside is reachable rather than merely rhetorical is timing. Public evidence points to 22one being formed in late 2025 and reportedly reaching an approximately $400 million valuation within roughly a year, while still in stealth [Dealroom News, October 2026] [Financial Times, October 2026]. That is a thin factual base, and it should be treated carefully, but it implies the market is already assigning option value to a policing-AI platform thesis, not just to a single feature. If 22one is building software that sits close to investigation workflows, redaction, intelligence handling, or evidence review, the addressable value per account could move well beyond departmental tooling into infrastructure budgets over time, especially if deployment risk can be contained [The Hustle, November 2026] [LinkedIn].
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Investigations system of record | 22one launches from stealth with workflow software that becomes embedded in core investigative processes across UK and allied police forces, then expands account scope from one use case into multi-team deployment | A named force deployment or formal launch that shows the product sits inside day-to-day investigative work rather than at the edge | Public reporting already frames the company around improving police efficiency, and Richardson's operating background gives the company unusual buyer fluency for this market [Inc., October 2026] [Dealroom News, October 2026] [Reuters, February 2008] |
| Trusted AI compliance layer for policing | 22one wins around a narrow, high-need function such as document handling, redaction, intelligence review, or case preparation, then becomes the approved AI layer for sensitive law-enforcement data workflows | A first product release that solves a compliance-heavy pain point and shows lower operational risk than general-purpose AI tools | Richardson previously co-founded Riven, whose DocDefender automated redaction for police documents and later won an NPCC Commissioner's Award with Bedfordshire Police, which supports a repeatable wedge in narrow but important workflows [LinkedIn] [Dealroom News, October 2026] |
| Public-safety platform for allied agencies | The company uses police credibility to expand into adjacent public-safety and justice buyers, creating a broader govtech platform across investigations, intelligence, and public-safety operations | A partnership, procurement vehicle, or reference deployment that makes expansion outside frontline policing easier | Richardson's public résumé includes policing leadership plus roles at AWS and Palantir in public safety, a combination that can translate beyond a single agency type if the core product works [RocketReach] [Liverpool Echo] |
The compounding mechanism here would be trust first, workflow depth second, and data advantage third. In this category, one credible deployment can do more than add one logo. It can shorten procurement cycles, improve security and policy credibility, and create a library of validated use cases that travel from one force to another. If the initial wedge is document-heavy or investigation-heavy, product usage can also generate institution-specific templates, review patterns, and workflow integrations that make replacement harder over time, even without a consumer-style network effect [Inc., October 2026] [Dealroom News, October 2026].
There is at least a faint public signal that this flywheel could exist. Richardson's earlier work at Riven was tied to a concrete police workflow, automated redaction, and an award linked to Bedfordshire Police and the National Police Chiefs' Council ecosystem [LinkedIn]. That does not verify traction at 22one, and it should not be read as carryover demand. It does suggest the founder has previously worked on a narrow policing problem where software value can spread through referenceability rather than mass-market distribution. In govtech, that is often how compounding begins.
The size of the win is harder to anchor because public reporting on 22one is thin and no direct public comparable is established in the source set. Even so, the current reported valuation, approximately $400 million as of October 2026, provides one concrete starting point [Financial Times, October 2026]. If the "investigations system of record" scenario plays out and 22one becomes a standard software layer for policing workflows across multiple jurisdictions, the upside case is not a modest step above that mark but a multibillion-dollar company, well above its latest reported valuation (scenario, not a forecast). The reason is simple: infrastructure software for regulated, high-trust public-sector workflows can command durable budgets and low churn once embedded, and the investor syndicate appears to be underwriting that category-scale possibility before product launch [EUVC, October 2026] [Inc., October 2026].
Lightly corroborated -- This section relies primarily on Financial Times-backed valuation reporting, secondary coverage from Inc. and Dealroom News, and public profile evidence for the founder's background. Product scope, customer adoption, and the specific compounding mechanism remain only partially corroborated in public sources.
Sources
Public sources
[22one] 22one | https://22one.ai/
[Dealroom News, October 2026] Ex-detective's stealth startup 22one hits ~$400M valuation, backed by Sequoia and Lux | https://dealroom.co/news/167125-ex-detectives-stealth-startup-22one-hits-400m-valuation-backed-by-sequoi/
[Inc., October 2026] A Former Detective's Startup Is Valued at Nearly $400 Million. It Still Hasn't Shown Its Product | https://www.inc.com/liyana-illyas/former-detective-startup-valued-at-nearly-400-million-dollars-still-has-not-shown-product/91426801
[Financial Times] Ex-Merseyside police officer lands $400mn valuation for AI ... | https://www.ft.com/content/65a113b4-2a45-49cf-9e68-93f42151166e?syn-25a6b1a6=1
[Reuters, February 2008] Police smash drug gang in huge raids | Reuters | https://www.reuters.com/article/uk-britain-raids/police-smash-drug-gang-in-huge-raids-idUKL1391124320080213/
[RocketReach, Retrieved 2026] Steven Richardson | RocketReach | https://rocketreach.co/
[Liverpool Echo, Retrieved 2026] Steven Richardson held roles of head of policing and public safety and principal advisor of global public safety at Palantir Technologies | https://www.liverpoolecho.co.uk/
[The Hustle, November 2026] Clearing the air | https://thehustle.co/newsletters/02-11-2026
[EUVC, October 2026] 22one raises undisclosed round at near-$400m valuation | https://www.eu.vc/p/22one-undisclosed-round-police-ai
[LinkedIn, Retrieved 2026] Anders Berntsson - Lovable | LinkedIn | https://www.linkedin.com/in/anders-berntsson-60ab5174/
[Tech Funding News, October 2026] Report: Ex-detective’s stealth startup 22one valued at nearly $400M, backed by Sequoia and Lux | TFN | https://techfundingnews.com/22one-stealth-police-ai-startup-400m-valuation-sequoia-lux/
Articles about 22one
- Sequoia and General Catalyst Back a Police AI Startup in Stealth — Backed by Sequoia and General Catalyst, the London startup is building AI tools for law enforcement before showing its product.