Margin of Safety #24: Black Hat 2025

In a break from our usual blogging, we spent last week at Black Hat checking out the latest and greatest security offerings. Our thoughts below!

· Exit chatter among AI security startups
Right before RSA, Prompt Security announced its sale to SentinelOne for $250M. That’s a great outcome for Prompt and their investors. At Black Hat, we heard many rumors and chatters about other early-stage AI security vendors in talks for M&A and many were simultaneously looking for large rounds (dual tracking). By contrast, some of the public equity investors around SentinelOne expressed frustration over the prospect of a large strategic acquisition, and the subsequent dip in stock price.

· AI-powered vulnerability management dominates
A clear wave of “AI vuln management” solutions is emerging: platforms that ingest traditional vulnerability scans, apply deep contextual analysis to prioritize critical findings, and dramatically cut alert noise. Unfortunately, we think this is at most half of the battle. As CVE exploit times continue to fall, part the challenge operators face isn’t just noise but rather an increase in patch frequency to stay ahead of the exploit curve. This drives an increase in raw work that teams are not necessarily resourced for, leading us to the second requirement:

· Prioritization alone won’t be enough; costs must also come down
While AI-driven prioritization can solve part of the problem, the community is still searching for reliable solutions that not only flag vulnerabilities but reduce the pure cost of remediating them at scale. We’re particularly interested in tools to help with via safe, incremental micro-patches or that reduce the extended effort surrounding patch rollout – things like checking (or improving) test coverage, wrangling staging environments, and managing actual deployments.

· Continued low satisfaction of current AppSec solutions

CISOs use multiple (5+ appsec vendors) that often overlap, but still do not have the outcome that they want. CISOs are really interested in finding solution that their small appsec teams can use to augment their capabilities. Their appsec teams are underwater and they know it, but they haven’t yet found a silver bullet solution. This relates to our previous hypothesis: we suspect that any such bullet will need to move beyond prioritization and into fixing.

· Everything is an agent now

Most booths claimed to have an agentic solution, though a decent fraction didn’t seem sure what constitutes an agentic solution anyways. We (or at least Kathryn) hope that this forces the market to start communicating about the practical value of their solution instead of tossing around marketing buzzwords! Alas, it’s more likely that a new buzzword will emerge.

· Phishing is top of mind
We cannot find anyone who disagrees that phishing is increasing in both volume and efficacy – this probably means this bullet is news to no one, but we’re still going to call out that this has implications up and down the stack for providers who secure human interactions.

· Rise in per-company AI models
As everyone starts to roll out an agent, is the next marketing hook going to be a model specifically tuned for your environment? Our tech cynic questions whether this is strictly necessary for all use cases, but it anecdotally feels like attack paths and local context are on the rise as marketing catchphrases.

· Well-financed newcomers at the booths
Startups fresh off sizable funding rounds (e.g. XBOW) made strong floor presences, paying the steep cost of prime Black Hat booth real estate. We noticed many startups opting for more affordable B-Sides showcases. The theme of Black Hat is the new RSA and B-Sides is the new Black Hat has some analogies in the AI conference landscape as well. Folks who are cultivating low business, highly technical conferences may need to be thinking about how to keep them that way!

Reach out to us if you are building in this space. We have some thoughts!
Kathryn Shih – kshih@forgepointcap.com
Jimmy Park – jpark@forgepointcap.com

This blog is also published on Margin of Safety, Jimmy and Kathryn’s Substack, as they research the practical sides of security + AI so you don’t have to.