Humanos raises $3.2M seed to give AI agents a risk score
Anthemis leads a $3.2M seed into Lisbon's Humanos, whose 0-100 risk scores and risk passports aim to let insurers and lenders underwrite autonomous AI agents - 350+ deployments already.
Also see more funding coverage: /funding/
Founded 2025 · Lisbon, Portugal
Lisbon-based Humanos has raised $3.2 million in seed funding led by Anthemis, joined by Springbank, Critical Ventures, Start Ventures and the founders of Sword Health, Feedzai and OutSystems, to build the live risk scores that let insurers and lenders underwrite autonomous AI agents.
The round, announced on 28 September 2026, lands on a question most of the industry has been happy to postpone: who carries the risk when an AI agent acts on its own? Agents are already booking travel, moving money and touching patient data – and the moment one of them gets it wrong, there is no policy, no credit line and no actuarial table waiting. Humanos wants to be the company that makes those instruments possible.
The new capital goes into expanding the company’s risk datasets, hardening its core infrastructure and growing the network of insurers and credit providers that plug into the platform, per the announcement. The team has set itself a public target too: $1 billion in agentic economic activity processed through the platform within the next 12 months.
A credit score for software that acts
Humanos calls itself the risk network for AI. The product assigns each AI agent a continuously updated score from 0 to 100, computed from the agent’s behaviour, permissions and activity history, and writes the trail into what the company calls a risk passport. Insurers, lenders and the businesses deploying agents all read the same passport – and the score moves with every action, because each action changes the context the next one runs in.
Founded in 2025 and led by co-founder and CEO Pedro Andrade, the company reports more than 350 deployments across fintechs, hospitals and insurers, per the announcement – InsureNow, Agentics Credit and Numo among the named customers, alongside work with Portuguese hospitals, Lusíadas included, on running patient consent collection through AI agents. For a company barely a year old, the customer list is doing most of the arguing.
The trust layer is a $14 billion market
The segment Humanos sells into – AI trust, risk and security management – is forecast to grow from $3.4 billion in 2026 to $14.1 billion by 2033, a 22.6% CAGR over 2026-2033 (Grand View Research, August 2026). That is the growth rate of a problem nobody has solved yet: enterprises are deploying agents faster than they can govern them, and the governance spend follows with a lag.
The cheque sits exactly where our own fundraising data says this layer is being built. In the same September week, Munich’s Kontext raised a $4 million seed to police what AI agents do at runtime, and Paris-based Primo raised €6.9 million to send agents into corporate IT. Humanos’ $3.2 million is the smallest of the three cheques – and arguably the most structural, because insurance is what turns a pilot into a rollout.
Lisbon is building the load-bearing layer
Now the pattern. Europe did not build the frontier models – we have all read that column. But look at what this month’s seed rounds actually funded: the security that polices agents, the operations they run, and now the rails that price their risk. The agent economy is being assembled in layers, and European founders keep claiming the load-bearing ones. The good news is that these are the layers with recurring revenue attached – every enterprise that deploys an agent eventually asks who insures it. For scale, San Francisco’s AIUC raised a $15 million seed for AI agent insurance in July 2025 (Fortune, July 2025): the US is a year ahead on capital here, not on the idea.
The syndicate is the other tell. When the founders of Sword Health, Feedzai and OutSystems – three of Portugal’s biggest tech stories – back the same Lisbon seed, that is a flywheel turning: exits becoming angels becoming the next generation’s first believers. Feedzai spent a decade scoring the risk of human transactions; its founders backing a company that scores machine ones is as clean a torch-pass as this ecosystem writes.
What to watch over the next year is not the score – it is whether an insurer writes a real policy against a risk passport, and whether that $1 billion of processed agentic activity actually shows up. If it does, Lisbon will have quietly built the trust layer the agent economy runs on. Watch the passports, not the demos.
