What Just Changed

On May 17, 2026, Linus Torvalds posted his Linux 7.1-rc4 kernel announcement — and buried in it was a direct, unambiguous warning, reported by The Verge and Slashdot. The Linux security mailing list has been rendered "almost entirely unmanageable" by a flood of AI-generated bug reports. His words, not ours.

Simultaneously, the Linux kernel project merged updated "security-bugs" documentation that formally rewrites the rules for how AI-assisted reports must be handled.

What Torvalds Actually Said

Torvalds was blunt, per Slashdot's direct quotes from the kernel mailing list:

> "The continued flood of AI reports has basically made the security list almost entirely unmanageable, with enormous duplication due to different people finding the same things with the same tools."

The core problem: people running identical AI scanning tools are each submitting the same bugs — independently, privately, clogging the queue. Maintainers spend their days forwarding duplicates or typing "that was fixed a month ago" instead of writing actual code.

Torvalds called this "entirely pointless churn."

He added a direct call-out to drive-by reporters: "If you found a bug using AI tools, the chances are somebody else found it too. If you actually want to add value, read the documentation, create a patch too, and add some real value on top of what the AI did."

The New Rules — What Actually Changed in the Documentation

The Verge and Cybersecurity News both covered Torvalds' frustration, but the actual policy changes in the kernel documentation carry more weight than the quote alone.

According to Cybersecurity News, the new Linux kernel security documentation now:

• Reserves the private security list exclusively for urgent, easily exploitable bugs crossing a clear trust boundary on properly configured production systems.
• Explicitly classifies AI-detected bugs as public by default, because they "systematically surface simultaneously across multiple researchers, often on the same day."
• Requires reporters to actually reproduce the issue, include a tested reproducer, and ideally submit a patch — not just fire off a tool-generated report.
• Bans heavy formatting and speculative "what if" chains — reports must be plain text, concise, and focused on verifiable impact.

If your AI tool found it, treat it as public. Don't waste the private security list's time.

This Is Bigger Than Linux

Torvalds' complaint landed the same week Ars Technica reported that bug bounty programs across the industry are drowning in the same flood.

Bugcrowd — whose clients include OpenAI, T-Mobile, and Motorola — saw reports quadruple in a three-week period in March 2026, according to Ars Technica. Most were false.

Curl's creator Daniel Stenberg suspended Curl's paid bug bounty program in January after what he called an "explosion in AI slop reports." He wrote in a blog post that the "never-ending slop" had taken "a serious mental toll to manage."

Nextcloud suspended its bug bounty program in April due to a "massive increase of low-quality reports," according to Ars Technica. It's hoping to restart once it figures out how to filter submissions.

Sophos CISO Ross McKerchar told Ars Technica there are now three categories of bad actors: amateurs trying AI for the first time, experienced researchers getting "led on" by AI agents, and a third group of "experienced AI builders" who built automated end-to-end scanning-and-submission systems creating "absolute carnage."

GitHub senior product security engineer Jarom Brown also weighed in, per The Verge: GitHub has "no problem" with AI tools in general, but AI-assisted reports must be validated to be useful. An unverified AI finding with no proof of concept is worthless.

What Mainstream Coverage Is Missing

Most outlets covered Torvalds' quote and moved on. A few key connections:

Anthropic launched Mythos last month — its dedicated cyber AI model that it claims can find software flaws faster than humans, according to Ars Technica. Purpose-built AI security tools entering the market means the volume of automated submissions is going to accelerate. The documentation Torvalds just merged is a preemptive adaptation to a problem that's still growing.

The economic distortion also matters. Google's bug bounty program paid out $17 million in 2025, up from $7.5 million in 2021, per Ars Technica. Six-figure payouts exist. That's a financial incentive powerful enough to drive mass automation. People aren't spamming bug reports because they love chaos — they're doing it because the payout structure rewards volume over quality.

The Current Threat

The open-source security infrastructure that underpins most of the internet — Linux runs on servers, phones, routers, and government systems — is now being taxed by people who downloaded an AI tool and hit "scan" without understanding what they were looking at.

Torvalds fixed it the only way you can: formalize the rules, remove the financial incentive for lazy submissions by stripping them of "private" status, and tell people bluntly to either add real value or get out.

The bug bounty industry has no equivalent gatekeeper. Several programs are already suspended. If the industry doesn't build its own triage infrastructure fast, the entire bug bounty model — which has paid out hundreds of millions of dollars in legitimate security discoveries — starts to collapse under the weight of AI-generated noise.

Curl and Nextcloud have already shut down their programs. The clock is running.