Post-Cutoff

Policy & safetyAnthropic and Horizon3.ai92 days after June 2026

Mythos-found Rejetto HFS auth bypass is exploited in the wild a day after disclosure

Confirmed

The takeaway

On Sept 30, 2026 Horizon3.ai disclosed CVE-2026-61500, a critical authentication bypass leading to RCE in Rejetto HTTP File Server 3.x, found by a Horizon3 analysis agent running Claude Mythos.

Status
Claim

Confirmed

Our reporting
High confidence
Importance
2 of 5
Last verified
4 October 2026

Your AI and this story

  • GPT-6.2 Astra122 days after its cutoff
  • Claude Opus 5.592 days after its cutoff
  • Gemini 4 Hydrogen183 days after its cutoff
  • Grok 4.7122 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 92 days before it.

Key facts

  • Disclosure: Horizon3.ai (Zach Hanley), Sept 30, 2026; affects Rejetto HFS 3.x (TypeScript rewrite); The Register says upgrade to v3.2.1 or later
  • Mythos noted that ‘consecutive doubles allow xorshift128+ state recovery (standard Z3/algebraic tooling)’ and that the app leaked raw Math.random() outputs, so session cookies could be forged for admin access
  • Horizon3: its human researchers chose targets and validated findings, and said that without the model they would likely have dropped this kind of cryptographic bug for lack of time and maths expertise
  • Exploitation from Thursday evening (Oct 1), first from a China-hosted IP against US and Japanese servers, then from US proxy IPs (The Register, citing VulnCheck’s Patrick Garrity)
  • As of Oct 2, Mythos and Project Glasswing were credited with 286 CVEs, and only one had previously been exploited in real-world attacks (The Register)

What happened

A security firm using Mythos through its own analysis harness found a cryptographic bug chain in a popular small file server, published it with a fix available, and attackers began exploiting it within about a day.

Why it matters

It shows both sides of AI vulnerability discovery: the model found a multi-step maths-heavy bug that humans said they would likely have skipped, and the gap between disclosure and exploitation was about a day.

Sources

2 sources from 2 sites. Numbers match the chips in the text.

2 sources: 1 primary, 1 press

Primary

  1. Horizon3.ai: Anthropic Mythos discovers Rejetto HFS RCE (CVE-2026-61500)horizon3.ai, official

Press

  1. The Register: Anthropic’s super bug-hunting model Mythos is hardcore good at math, as latest vuln under attack shows (Oct 3)theregister.com, press

Changes

  • Filed

Status

Claim

Confirmed

Our reporting
High confidence
Importance
2 of 5
Last verified
4 October 2026

Sources at a glance

2 sources: 1 primary, 1 press

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
4 October 2026
Human review
None recorded for this entry. What the editor does
Version
Last saved 5 October 2026

Spotted an error? Write to contact@postcutoff.com. Corrections are logged in public.

This page for your AI

Same text, no layout:

Open in ClaudeOpen in ChatGPT

Related

On “Should we be worried?”

  • Context Epoch/Ipsos polls found no rise in US adults reporting cyber incidents since Fable 5 (46.2% to 44.6%); only 2 of 286 Mythos/Glasswing CVEs were known to be exploited.In Cyber capabilities and misuse

Related events

  1. Model releases

    Anthropic releases Claude Fable 5.1 and Claude Mythos 5.1

    Confirmed

  2. Model releases

    Anthropic reveals Claude Mythos Preview, withholds it over cyber risk and launches Project Glasswing

    Confirmed

The whole topic