Essay argues cheap AI agents will make cyberattacks routine
Original title: Cybersecurity Is About to Be the Biggest Problem in Tech
It cites an authorized NSA Mythos test, a claimed Cairn campaign, and slower patching times.
A Substack essay argues that inexpensive autonomous agents will make opportunistic attacks routine, especially against smaller organizations. It recounts an authorized NSA red-team test of Anthropic’s Mythos and a claimed Cairn campaign against more than 1,200 websites. It also cites patch-timing figures, the JADEPUFFER ransomware case, and FBI warnings about voice cloning. These are the author’s examples rather than new confirmed incident findings.
Unconfirmed: Data theft in the claimed Cairn campaign is unknown, and much of the piece is opinion.
- The NSA Mythos exercise was an authorized test, not a real breach.
- A researcher claims Cairn agents entered sites of more than 100 companies.
- Verizon’s 2026 report is cited for ransomware in 83% of small-org breaches.
- The essay says median critical-bug patch time has risen to 43 days.
- KnowBe4 is cited for a 41% rise in Microsoft Teams attacks.
- Vendors
- Anthropic
- Malware
- JADEPUFFER
- Victims
- DeereSharkNinjaCummins
- Countries
- United StatesChina
- AI models
- MythosDeepSeek V4 Flash
From the original article
11 min read at projectautonomy.substack.com
Back in June, the head of the NSA told a U. S. senator that Anthropic’s Mythos model broke into “almost all of [the NSA’s] classified systems, not in weeks, but in hours.”
By projectautonomy.substack.com by lukewilson · projectautonomy.substack.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.