Interview with Douglas Day and Michiel Prins
AI Is Not Killing Bug Bounty: It Is Raising the Bar
AI is making it easier for more people to attempt vulnerability research, but Douglas’ experience shows that the biggest gains are going to researchers who already know how to find meaningful bugs. For elite security researchers, AI is not a replacement; it is a force multiplier. There is growing concern that AI will either replace security researchers or flood bug bounty programs with poor-quality reports. Douglas and Michiel can offer a more balanced view: AI can help experienced researchers scale their instincts, but it does not replace the judgment, context and impact assessment that make a report valuable.
Topic Segment: Can an AI agent ever truly have an identity?
(from Mike Privette's Security, Funded Newsletter)
- 60% - Yes, and it can be separate from humans
- 10% - Yes, but it only assumes the identity it runs under
- 10% - Yes, by use of whatever credentials it finds
- 10% - No, everything is permitted unless explicitly denied
- 10% - Other (tell me)
Mike’s Thoughts:
Most of you think an AI agent can have an identity of its own. A good chunk of the "Other" crowd showed up to argue about what "identity" even means in the first place, which honestly might be the better question. The market already seems to agree agent identity matters, just not as its own category. Over the past year, the biggest identity deals went to platforms outside traditional IAM, like Cyera buying Oasis, CrowdStrike buying SGNL, and Cisco buying Astrix. My bet is that agent identity ends up as a feature of whatever platform the agent already runs on, not something you buy on its own.
Weekly Enterprise News
Finally, in the enterprise security news,
- a little bit of funding
- new supply chain attacks
- new AI attack tools, like ARTEX
- OpenAI builds a kill switch to save us from the AI apocalypse that they might be directly responsible for
- A cybersecurity vendor takes their cloud platform offline after an angry researcher goes full disclosure on them
- Glasswing seems to be wrapping up with a whimper rather than a roar
- Lessons from the incident response trenches
- Satellites run doom
- and a company named “Who Gives a Crap” gets breached
All that and more, on this episode of Enterprise Security Weekly.
Michiel Prins is a Co-Founder and Senior Director of Product at HackerOne. He is an information security expert, researcher, hacker, and developer. Michiel has been finding critical software vulnerabilities in technology for over 10 years. Prior to founding HackerOne, Michiel co-founded a successful penetration testing company that worked on projects for trusted organizations from government institutions to top technology companies, including Twitter, Facebook, Evernote, and Airbnb, among others. Michiel regularly presents on vulnerability disclosure and security research projects regarding security management, privacy, and web application infrastructure. Michiel graduated with a B.S. in Computer Science from Hanze University Groningen.
Douglas Day, known as Archangel, is an elite security researcher and top-ranked HackerOne researcher based on the US West Coast. Active on HackerOne, Bugcrowd, and other bug bounty platforms, he has been hacking since 2018, when a talk at BSides Portland pulled him into the field. In 2026 he moved from fully manual hacking to AI-driven automation, building Skitarii, a custom bug bounty pipeline that triages findings and files reports through the HackerOne API.
The hardest part of defending an enterprise today isn't finding security gaps. It's managing complexity.
Hybrid infrastructure. Legacy systems. Cloud. AI. Third-party risk. Regulatory requirements. Nation-state threats. And somehow you're expected to make all of it work together.
If you're responsible for securing government or critical infrastructure environments, join us for the Government Security Virtual Summit on November 18. Hear how security leaders are modernizing defenses, reducing complexity, and building resilient security programs that stand up to today's threat landscape.
Security Weekly listeners can register for free at https://securityweekly.com/govsec using the discount code CSS26-SW
Adrian Sanabria
- FUNDING/M&A, courtesy of the Security, Funded newsletter, issue #269 – Fiddler on the [Falcon]
FUNDING
- Sentra (IL) · $150.0M Series C | cloud-native data security platform, led by Capital One Ventures
- Lazarus AI · $43.0M Venture Round | AI source-code vulnerability platform, from QBE Ventures, Mucker Capital, Motivate Venture Capital, State Farm Ventures, Wave Function Ventures, Indelible Solutions, Joyance Partners, Allegis Capital, One Way Ventures, AngelList, Techstars, and Plug and Play
- Brava Security (US) · $15.0M Seed | cloud security telemetry engine, led by F2 Venture Capital and Pitango VC
ACQUISITIONS
- Dreadnode (US) · adversarial AI simulation and red teaming | acquisition by CrowdStrike, undisclosed
- Fiddler AI (US) · AI model and agent observability | acquisition by CrowdStrike, undisclosed
- ATTACKS: New Supply Chain Attack Compromises Tens of Thousands of GitHub Repos
- ATTACKS: ARTEX: Agentic Asbestos – Censys
ARTEX, one of many AI-based pentesting framework/harness that have popped up in the past weeks and months, is causing some widespread chaos.
- ANTI-APOCALYPSE: OpenAI builds a kill switch for its research models
- VULNERABILITIES: N-able takes cloud platform offline after 30+ “zero-days” published
The more troubling part of this story is the source of the chaos: a researcher that works as a penetration tester and holds many cybersecurity certifications, including a CISSP. Most folks in cybersecurity understand that businesses have to balance priorities and cybersecurity isn’t always the top concern. Some, however, see security as a binary situation - if a vulnerability is found, it must be fixed.
When I advise IANS clients, I warn that a portion of Good Samaritans are the most dangerous of potential bug reporters. When they find a serious vulnerability, they won’t just report it and forget about it. They need to be taken seriously and given explicit expectations of what is going to happen. You must close the loop with them. If they don’t think you’re taking a bug they’ve submitted seriously, they will often publicly disclose it in an effort to use the media to pressure the company into fixing the issue.
Sometimes the bug is a false positive and the researcher refuses to believe it. Sometimes it’s a legit, critical vulnerability, and the vendor is in the wrong. One bad experience is often all it takes - once trust is lost, the researcher defaults to public (aka “full”) disclosure without giving any prior warning to the vendor or its customers.
In this particular case, the researcher had a bad experience reporting vulnerabilities to N-Able earlier this year and felt ignored, as they still have not been fixed. He used AI to find dozens more vulnerabilities and went the full disclosure route. He may feel like he’s punishing the vendor, but all the vendor’s customers are also impacted. The vendor blames the researcher for “non-responsible communication”, but in the researcher’s eyes, the vendor is the one shirking it’s responsibilities.
Regardless of who should bear the blame for the situation, the vendor is not entitled to coordinated disclosure (as far as I’m aware, I’m not a lawyer). However, the researcher has a CISSP and his behavior seems like it would violate the ISC2 code of ethics. He also lost his job in the last 24 hours, which could be a coincidence, but might not be - it could be difficult finding a job after publicly forcing a vendor to take their systems offline. Similarly, I don’t think I’d recommend anyone use products from this vendor. Trust goes both ways and I don’t think either party came out of this situation looking great.
- VULNERABILITIES: After 7-months, Glasswing Overwelms Maintainers with Underwhelming Findings
- VULNERABILITIES: *sshole · sshhole.com
Boy, this one burned bright and fizzled fast.
- LESSONS LEARNED: Black October: Lessons From the Downstream Breaches
- HACKS: CHIRP // AUDIO DATA TRANSMISSION
Transfer data with sound!
This isn't the first time we've seen a sound-based side channel attack: singing pixels
- INCIDENT RESPONSE: Cybersecurity firms, overwhelmed by AI attacks, are turning away new clients
- SQUIRREL: Who Gives A Crap data breach raises questions over third-party security risks
It was a real shitshow.
- SQUIRREL: Yes, it runs ‘Doom’: Satellite plays iconic game in world’s 1st demonstration of orbital voice control
Sean Metcalf
