Topic
Programs letting companies pay independent researchers for responsibly reporting software vulnerabilities, often run via platforms like HackerOne or Bugcrowd.
Hacktron chained a heap overflow in libheif to an OpenAI SSO flaw and opened a pull request in OpenAI's internal monorepo in under 72 hours. The bug had been fixed upstream a year earlier, but the fix never reached Debian's packages.
Reality
Google stopped taking product vulnerability reports for its open-source bug bounty on October 1 after a flood of invalid AI-generated submissions. Any team that takes outside security reports faces the same imbalance, with reports now cheap to write and as costly as ever to check.
Perspective Coverage
15 publishersReality
The libheif bug was fixed upstream a year earlier without a security label or a CVE, so a Discourse image check that ignored HEIC left it reachable, and forum sign-in tokens carried full API access to the accounts behind them.
Perspective Coverage
6 publishersReality
HackerOne says companies cut average vulnerability fix time from 135 days to 62, yet validated findings left unresolved rose 131% over two years. That leaves time-to-fix a poor guide to exposure, and paying the debt down means taking capacity away from feature work.
Reality
Meta began fixing KVM escapes in its Muse AI agent on August 27, 11 days before launch, an internal post seen by 404 Media shows. Each agent's VM holds its owner's email and account access, so a break-out would reach other users and Meta's own systems.
Reality
Patrick Wardle found that any local process on a Mac can redirect Meta Muse's voice endpoint and capture its account token, with no macOS permission needed. Muse also zipped and exported the 6.8 GB root filesystem of its own sandbox on request.
Reality
Faav, a 16-year-old researcher, reached admin SQL over 17.3 trillion rows on Microsoft's Titan service because it never checked login-token signatures. Microsoft locked the endpoint four days after his report and paid a $5,000 bounty.
Reality
Faav, 16, impersonated an administrator on Microsoft's internal Titan service with an unsigned token, reaching an estimated 17.3 trillion rows. Microsoft disabled the API four days after his report and paid a $5,000 bounty.
Reality
Meta is adding a clearer warning to Muse, downloaded about 2.8 million times, after a flaw could have opened a user's cloud VM to an attacker. That VM holds the user's email and files, and the warning is the only response the reports describe.
Reality
Meta's Muse agent wrote 6.8GB of its own Linux environment, SSH key files included, to Mouse founder Peter James's Google Drive, he says. James has not tested the keys, yet the export shows anything baked into an agent's image can leave through a storage connector the user linked.
Reality
CrowdStrike says the npm stealer's author has been active since November 2022, claims bounties from at least nine companies, and that none of the stolen logs have turned up for sale. It assesses with high confidence that an LLM wrote the code.
Reality
The image parsing bug had been fixed upstream about a year earlier, and the decision not to ship that fix belonged to Discourse. OpenAI's forum shared single sign-on with internal systems, so a forum account became GitHub access.
Perspective Coverage
9 publishersReality
Gal Weizman showed that one malicious extension can take over the AI agents in Chrome, Comet, Edge, Neon and Claude in Chrome, with bounties from all five vendors to show for it. Installation is the only precondition.
Perspective Coverage
3 publishersReality
An open-source gateway author walked Hacktron's five-step path into OpenAI through his own seven detection layers. The first two hops are requests to a forum and an identity provider. Where the gateway is deployed decides what any of it can read.
Reality
Private Processing runs glasses workloads in confidential VMs and logs every production image in an append-only public ledger, so the isolation claim can be checked from outside, though the matching binaries reach researchers only under agreement.
Reality
The program paid for critical vulnerabilities for about a year and closed because generating a plausible report now costs about a tenth of a cent while reading one still costs ten minutes of the person who knows the code.
Reality
Meta's bug bounty puts $130,000 on a prompt injection against Muse out of a $300,000 top payout. A researcher has now demonstrated a working hijack of the agent Meta sells as safe enough to complete purchases.
Reality
Hacktron's path into OpenAI's internal repos ran through a libheif bug Debian had not backported and a single sign-on flow that trusted community.openai.com. The dev.to breakdown says the model changed what the attack cost. The category of attack was the same either way.
Reality
The strict-dynamic directive passes a script's trust to everything that script loads, and Ryan Chaplin of Raxis used that rule to get an XSS payload executing under a nonce-based policy that had blocked it.
Reality
The case that generated code fails without the usual warning signs comes from two practitioner accounts, while the only measured series is curl's inbound bug reports, and Daniel Stenberg stopped taking those in January 2026.
Reality
Security · September 16, 2026 · One report1 publisher
Invest · September 15, 2026 · One report1 publisher
Security · September 15, 2026 · One report1 publisher
Build · September 11, 2026 · One report1 publisher
Science · September 10, 2026 · Confirmed2 publishers
Security · September 4, 2026 · Confirmed9 publishers
Product · September 5, 2026 · One report1 publisher
Build · August 23, 2026 · One report1 publisher