Matilda Bailey is a seasoned authority in the high-stakes world of networking and next-gen security solutions, having spent the last decade navigating the shift from manual hardware audits to the current era of hyper-automation. As we stand in 2026, her perspective is more critical than ever, particularly as the industry grapples with the fallout of the generative AI boom that has fundamentally destabilized traditional vulnerability management. In this conversation, we delve into why global powerhouses like Intel and Google are retreating from the crowdsourced bounty models that once defined modern defense. We explore the tactical pivot toward internal AI auditing, the sheer exhaustion of engineering teams facing a deluge of “plausible-looking slop,” and what this permanent realignment of enterprise security budgets means for the future of the independent research community.
The security landscape has faced a massive upheaval recently due to the influx of AI-generated content. How is this “slop” fundamentally altering the way enterprises manage their vulnerability pipelines?
We are witnessing a structural collapse of the traditional open-door policy for vulnerability reporting, primarily because the signal-to-noise ratio has hit an all-time low. In early 2026, the volume of monthly disclosures skyrocketed, moving from 5,045 in January to a staggering 10,740 by August, a doubling of the workload in just over half a year. This isn’t just a quantitative problem; it’s an operational nightmare because these AI-generated reports often look credible at first glance, mimicking the language of sophisticated researchers while lacking any actual substance. It creates an asymmetric resource drain where a submitter can generate a professional-looking report in seconds for virtually no cost, while our internal teams must spend dozens of expensive engineering hours manually reproducing the finding to prove it’s a false positive. The sensory overload of triage teams is palpable, leading to a state of exhaustion where genuine, critical flaws risk being buried under thousands of hallucinations.
When we see long-standing pillars of the community, like the Curl project, essentially waving the white flag and shutting down their bounty programs, what does that reveal about the viability of current crowdsourcing models?
It’s a sobering moment that highlights the fragility of trust in the open-source ecosystem when faced with automated bad actors. Daniel Stenberg’s experience with Curl was a clear warning; he tracked a slow erosion starting in 2024 that turned into a full-blown crisis by 2025, proving that even the most resilient projects can be brought to their knees by a flood of LLM-generated junk. For a maintainer, the psychological toll of being forced to play the role of an unpaid editor for AI hallucinations is unsustainable, as it takes critical time away from actual development and patching. This signals that the “honeypot” effect of bug bounties—where you offer money to attract the best talent—has backfired by attracting a million low-effort scripts instead. We are reaching a point where the administrative overhead of a public program outweighs the security benefits, prompting a shift toward gated communities and invite-only research programs.
Intel and Google have both made headlines by pausing or pivoting their reward programs this year. What are the primary strategic drivers behind these major shifts from such influential tech giants?
The decision by Intel to pause its rewards and Google’s move to stop accepting reports for its OSS VRP on October 1st are purely pragmatic responses to a marketplace where the ROI has fundamentally shifted. These companies aren’t retreating from security; they are moving the battlefield inside their own walls where they can control the tools and the noise. By shifting budgets toward internal AI-driven auditing and “secure by design” frameworks, they can find and fix vulnerabilities more efficiently than any external crowd could, without the baggage of thousands of invalid reports. There is a growing realization that if you can use AI to generate a bug report, you can use it to build a better automated fuzzer or an agentic red teaming tool that operates 24/7. They are choosing to invest in their own proprietary automation rather than paying for the privilege of triaging someone else’s automated scripts.
As enterprises lean more heavily into internal auditing and agentic red teaming, how do you see the role and the “market value” of the independent bug hunter evolving?
We are entering a period of permanent realignment where the casual, “low-hanging fruit” hunter is effectively being phased out of the economy. Experts like Katie Moussouris have noted that bounty prices are unlikely to return to their previous highs because the perceived value of a basic vulnerability has plummeted in the face of internal automated testing. To stay relevant in 2026, independent researchers must transition from being “finders” to being “architects” who can demonstrate complex, multi-stage attack chains that AI still fails to conceptualize. The industry is moving away from paying for individual bugs and toward rewarding deep, high-fidelity human intelligence that can navigate logic flaws that an LLM simply cannot see. It’s a “flight to quality” where only the top 1% of researchers will continue to find lucrative opportunities, while the rest may find the barrier to entry becoming prohibitively high.
If organizations continue to restrict their programs to avoid the noise, is there a genuine risk that we will lose access to the “white hat” geniuses who find the most dangerous, systemic flaws?
That is the million-dollar question, and leaders like Kara Sprague are rightfully concerned that we might be throwing the baby out with the bathwater. When you stop the cash flow, you don’t just stop the “slop”; you risk alienating the brilliant researchers who have the intuition to find a “logic bomb” or a zero-day that no automated tool has even dreamed of yet. There is a palpable fear that these skilled individuals will take their talents to the private exploit market or, worse, to state-sponsored actors where the payouts remain astronomical and the triage process isn’t their problem. The industry has to get better at the work of sorting volume because these researchers are still worth listening to; otherwise, we are just creating a darker, more secretive market for the most dangerous vulnerabilities. We need to evolve our triage platforms to use AI as a shield that protects human reviewers, rather than using it as an excuse to shut the door on the community entirely.
What is your forecast for the future of vulnerability management over the next few years?
I expect that by 2028, the public-facing, “submit-and-get-paid” bug bounty will be a historical footnote for major corporations, replaced by a much more rigorous, credentialed ecosystem. We will likely see a move toward “Proof of Work” requirements for all submissions, where a researcher must provide a cryptographically verified, sandboxed execution of their exploit before a human even glances at it. Internal security budgets will continue to migrate—possibly upwards of 70%—toward autonomous patching and self-healing codebases that fix vulnerabilities in real-time as they are committed. The future is a “closed-loop” system where the only external input comes from highly vetted, elite security firms or long-term partners who are integrated directly into the development lifecycle. It will be a more resilient world, certainly, but it will be one that feels much smaller and more fortified than the open, collaborative web we once envisioned.
