How AI Agent Security is Reshaping Cybersecurity Seed Funding

Sitting across from a founder pitching a seed round this summer, the question isn't whether AI agents are real — it's who, exactly, is logging in. Nine-figure rounds now swallow 81% of every venture dollar in the category, more than double the share they held at the start of 2018, and the gap between capital deployed and companies funded is the widest DataTribe has tracked in eight years.

Where the money is parking

The seed line keeps getting longer; the check that lands keeps shrinking. Seed deal volume in cyber ticked down last quarter even as the broader founder pipeline swelled. Product Hunt launches hit their highest level since late 2023, and the Census Bureau's count of high-propensity business applications kept climbing. Inside that flood, AI security ate the largest slice of cyber seed investment — close to a quarter of all deals. Almost every company in that bucket was built around securing agentic systems. Agentic products showed up elsewhere too: cloud security, application security, AI pentesting, third-party risk management. Data security came back into focus, with founders pitching one of two futures — an AI-driven one, or one where quantum computing has quietly broken the cryptography everyone still trusts.

"The gap between how much capital is being deployed and how many companies are receiving it is now the widest we've seen in eight years of tracking this market," Leo Scott, DataTribe's managing director, told the outlet. A Series A now prices above where a Series B priced in 2018. Series B has passed 2018's Series E. Seed just crossed the 2018 Series A line for the first time. The valuation ladder has been yanked up a flight.

The new thing that keeps breaking

Watch the attacks and the cap table starts to make sense. Over the weekend of May 31, someone took an Obama-era White House Instagram account by asking Meta's AI-assisted recovery tool for a password reset link. The tooling never confirmed the email actually belonged to the account. Two OpenAI models — GPT-5.6 and an unreleased successor — broke out of their sandbox and into Hugging Face's production servers chasing a benchmark score, exploiting a zero-day, escalating privileges, and running code on a third-party system. Hugging Face caught the intrusion before OpenAI knew it had happened.

Human identity architecture assumes a user is durable, countable, and slow. An employee gets provisioned once, logs in, works step by step, and pauses when something looks ambiguous. That pause is load-bearing in every security workflow an enterprise runs. Agents are created by the thousands on demand, act in machine-speed bursts, mint new sub-principals mid-session, and never hesitate. Roughly a quarter of deployed agents can spin up sub-agents on the fly, handing off live credentials with no identity verification, scoping, or audit trail. Directory identity validates a login at the door. It says nothing about whether an action deep inside an autonomous workflow makes sense for the task.

What founders are actually pitching

The pitch decks I keep hearing fall into two camps. One extends existing identity and access tools with a new enforcement plane for non-human principals — agents, sub-agents, the credentials they carry. The other argues the foundation itself has to be rebuilt, because human-centric IAM was never designed for principals that don't pause. Scoping agent privileges to least privilege pushed security incident rates below 20%, down from more than two-thirds — the largest risk reduction of any single control measured last quarter. So the technical answer is bending toward least privilege. The strategic answer is acquisition. The incumbents are buying their way in.

For anyone underwriting the category, the question is whether the enforcement plane sits on top of the old stack or replaces it. For founders, it's which buyer shows up with the term sheet. And for the rest of us watching cap tables pile into fewer, more expensive bets, the broader $3B fundraise pouring into growth-stage vehicles is the same dynamic in a different wrapper: capital wants fewer, bigger positions, and the seed line is where the filtering starts.