$ today_in_10_seconds
The venture money is splitting cleanly into two poles: agents that do a valuable job, and agents (plus tooling) that police the agents. Zenity just raised $125M to secure enterprise AI agents — with SoftBank, Hitachi, and LG backing a bet that the real risk isn't a powerful model, it's what an agent does once it's set loose. Alongside a $35M raise for agentic threat-hunting and a $120M round for medical-billing agents, today makes one thing plain: securing the autonomous workforce has become its own fundable category.
$ lead_story
Zenity raises $125M to secure 'the era of 1 billion AI agents'
Zenity closed a $125M Series C led by Norwest — with SoftBank Vision Fund 2, Hitachi Ventures, LG Technology Ventures, and Qumra joining existing backers Intel Capital, Vertex, Third Point, and DTCP — bringing total funding to ~$185M. The company's thesis, as CEO Ben Kliger frames it, is a deliberate break from the last decade of AI security: earlier waves focused on protecting models and prompts, but Zenity puts the agent at the center — the visibility and governance enterprises need once an autonomous agent can take actions, make decisions, and touch critical business systems on its own. Its credibility rests partly on offense: Zenity Labs has disclosed real agent vulnerabilities (its AgentFlayer research, flaws in Microsoft Copilot Studio, attacks on Perplexity's Comet browser), and Gartner has called it 'the company to beat in AI agent governance.' For builders, the signal is that agent security is no longer a feature of the model stack — it's a distinct control layer that enterprises will buy separately, and one investors are now funding at scale.
// why_it_matters: The risk that enterprises are paying to contain isn't the model — it's what happens when an agent is set loose to act, and that's now a nine-figure market.
$ also_worth_attention
Mate raises $35M so agents can hunt the attackers
Mate raised a $35M Series A led by Canaan Partners to build a platform where trusted autonomous agents detect, respond to, and proactively hunt threats — arguing that traditional security operations simply weren't designed for AI-scale attacks. It's the flip side of Zenity: not securing your agents, but deploying agents as the defenders. The insight: defense is being forced to go agentic because the offense already has — when attackers move at machine speed, a human-paced SOC can't keep up, so the same autonomy that created the threat becomes the only thing fast enough to counter it.
link →Candid Health raises $120M — the other pole: agents that just do the job
Not every big agent round is about security. Candid Health raised a $120M Series D led by Sixth Street Growth to automate medical billing and claims processing with AI agents — the unglamorous, high-value back-office workflow that eats staff time and leaks revenue in every US healthcare provider. The insight: the funding market has two magnets, and this is the other one — vertical agents that own a specific, expensive workflow raise just as big as the security plays, because 'we automated a painful process with measurable ROI' is as investable as 'we'll keep your agents from going rogue.'
link →The pattern in the data: agent security is now the biggest slice of cyber seed funding
Today's rounds aren't isolated — they're the leading edge of a structural shift. Per Help Net Security, AI security was the single largest category of cybersecurity seed investment last quarter, close to a quarter of all deals, and seed round sizes crossed the 2018 Series A line for the first time. The forcing function is speed: CrowdStrike clocked 2026's fastest breakout — initial compromise to lateral movement — at 27 seconds, far faster than any human patch or threat-hunt cycle. The open question investors are underwriting: does securing non-human 'principals' bolt onto human identity tools, or require a new architectural foundation built for agents from the ground up.
link →
$ bigger_picture
The trend behind the headlines
Today's rounds map the shape of agentic AI's money in mid-2026, and it has two poles. At one, agents that do a valuable job — Candid Health automating medical billing — command nine-figure rounds because they attach autonomy to a specific workflow with measurable ROI. At the other, an entire industry is forming to secure and police those agents: Zenity's $125M to govern them, Mate's $35M to hunt threats with them, and a broader surge that made AI security the largest slice of cybersecurity seed funding last quarter. The two poles are causally linked — the faster enterprises deploy agents that touch critical systems, the more they need a control layer for agents that can act, err, or be hijacked at machine speed, especially when breakouts now happen in 27 seconds. The recurring lesson of the past month sits underneath it all: capability outran control, and capital is now funding both sides of that gap simultaneously. The unresolved question worth watching is architectural — whether agent security becomes a feature the big platforms absorb (the acquisition pattern is already heavy) or a durable new foundation built for non-human identity. Either way, 'who governs the agents' has graduated from a worry into a market.
Non-human identity (NHI) governance
The concept underneath today's security rounds and worth understanding even if you never buy a vendor. As agents proliferate, each one is a non-human 'principal' that authenticates, holds permissions, and acts — and traditional IAM was built for humans, not for entities that can spin up, delegate, and act autonomously at scale. The practical starting point for any team deploying agents: inventory your agents as first-class identities, scope each to least privilege, and log their actions for attribution — because you can't govern (or fund a defense for) agents you can't even see. It's the same 'know your agent' gap that keeps recurring across every agentic domain.
$ subscribe --free
Get the next edition in your inbox
// DAIly Agent covers agentic AI three times a week
// free, no spam, exit anytime