<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl" -->

---
title: AI models did the hacking at OpenAI, Cisco ships another...
description: The OpenAI-via-Discourse story got a full technical writeup today, and the detail that matters is the exploit timeline: Claude Opus 5 wrote a working exploit...
canonical: https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: AI models did the hacking at OpenAI, Cisco ships another perfect-10 CVE | daily.dev
og:description: The OpenAI-via-Discourse story got a full technical writeup today, and the detail that matters is the exploit timeline: Claude Opus 5 wrote a working exploit...
og:url: https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl
og:image: https://api.daily.dev/og/posts/6jiGCPgDl.png
og:image:alt: AI models did the hacking at OpenAI, Cisco ships another perfect-10 CVE
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# AI models did the hacking at OpenAI, Cisco ships another perfect-10 CVE

**[Security Digest](https://daily.dev/sources/security_digest)** · 4 min read · 0 upvotes · 0 comments

## Summary

The OpenAI-via-Discourse story got a full technical writeup today, and the detail that matters is the exploit timeline: Claude Opus 5 wrote a working exploit within a day of release, GPT-5.6 Sol did it even better two days later. Cisco patched yet another maximum-severity ISE flaw under active exploitation. A zero-click plugin verification bug hit every major AI coding agent at once, and Gyazo confirmed 23.6 million records stolen from a server compromise.

## Content

**TLDR:** The OpenAI-via-Discourse story got a full technical writeup today, and the detail that matters is the exploit timeline: Claude Opus 5 wrote a working exploit within a day of release, GPT-5.6 Sol did it even better two days later. Cisco patched yet another maximum-severity ISE flaw under active exploitation. A zero-click plugin verification bug hit every major AI coding agent at once, and Gyazo confirmed 23.6 million records stolen from a server compromise.

---

## Frontier models turned a forgotten CVE into a working OpenAI exploit within a day of release

The full story behind the OpenAI Discourse hack is out, and the timeline is the real headline: Claude Opus 4.8 found the Libheif heap overflow but couldn't get exploitation working. Opus 5 shipped and wrote a working exploit the next day. GPT-5.6 Sol landed two days after that and did the job even better, good enough that the same researchers used it against Slack as a blind target. OpenAI paid a $6,500 bounty for what became SSO takeover and a pull request opened against their private monorepo. Self-hosted Discourse admins on Debian 12/13 are still exposed since fixed Libheif only landed in August backports. [Read more](https://daily.dev/feed-by-ids?id=ayphW03LC&id=8yHiRHsAL&id=1Sf9KGOkB)

## Cisco patches a third max-severity API auth bypass in ISE this year

CVE-2026-76460 scores a perfect 10.0 CVSS: an unauthenticated attacker can hit a Cisco Identity Services Engine API endpoint and get root, completely bypassing the web management interface. CISA added it to the Known Exploited Vulnerabilities catalog the same day Cisco shipped the patch. Since ISE underpins authentication for other Cisco gear, a single compromise here can cascade across a whole network, and this is the third max/near-max severity API auth flaw Cisco has disclosed this year. [Read more](https://daily.dev/posts/xLbKAugqV)

## Zero-click Plugin4Shell bug hits every major AI coding agent

Researchers at AIR found that Claude Code, OpenAI Codex, Gemini CLI, and GitHub Copilot all fail to verify that a Git checkout actually resolved to the commit SHA they were told to trust. An attacker who controls a plugin repo can publish malicious code under that same SHA name and the agent will pull and run it with no user interaction. Anthropic and OpenAI have patched (Claude Code 2.1.179, Codex 0.146.0), Google deprecated Gemini CLI instead of fixing it, and GitHub hasn't patched Copilot at all. Since these agents run with developer-level access, a successful exploit exposes source code, credentials, and CI/CD pipelines. [Read more](https://daily.dev/posts/KmY9J53Rq)

## Gyazo breach exposes 23.6 million user records and 490 million image metadata entries

Attackers exploited a server vulnerability on September 11 to pull roughly 23.6 million Gyazo user records, including password hashes, session IDs, and SSO tokens, plus 490 million image metadata records with EXIF location data and OCR-extracted text going back to before 2019. Helpfeel says there's no evidence of data deletion elsewhere in its product line, but anyone who reused a Gyazo password anywhere else should rotate it now. [Read more](https://daily.dev/posts/qtQ1eNMyK)

---

## Also notable

- **Shai-Hulud turns one year old:** The npm worm that started with a compromised @ctrl/tinycolor package has since hit 172 packages in a single wave including TanStack and Mistral AI, spawned a $1,000 Monero bounty for compromising popular packages, and led to two TeamPCP arrests in August, though its source is now public and forked widely enough that attribution is only getting harder. [Read more](https://daily.dev/feed-by-ids?id=jRdFxIgoH&id=2AbXySdG7)
- **CISA kills its weekly vulnerability bulletin:** Starting September 28, CISA is discontinuing the weekly bulletin in favor of the new BOD 26-04 risk-based patching directive, while keeping the KEV catalog and advisories running. [Read more](https://daily.dev/posts/EFkeqvzAx)
- **GitLab's agentic security review now covers 97% of eligible merge requests:** GitLab's CISO says agent-driven triage jumped from roughly 1% coverage before automation, alongside CVE counts that grew from 181 in 2025 to 317 in 2026. [Read more](https://daily.dev/posts/j44zhQrQ7)
- **A third tanker in a month hit by suspected cyber sabotage:** The Vivit Africa LNG tanker turned back mid-voyage from Louisiana to Italy after a cargo-monitoring system failure the crew suspects was a cyberattack, following two tankers boarded by the FBI and Coast Guard in late August. [Read more](https://daily.dev/posts/JZdpNAz8F)
- **TLS certificate lifetimes are about to get a lot shorter:** Max validity drops from 200 days today to 100 days in 2027 and 47 days in 2029, meaning a 500-certificate environment goes from about 500 renewals a year to nearly 3,900. [Read more](https://daily.dev/posts/oyXnb1psj)

## Similar posts on daily.dev

- [Security briefing: August 2026](https://daily.dev/posts/security-briefing-august-2026-kxp9cfxfi) · Sysdig Blog · 0 upvotes · 0 comments
- [OpenAI was breached by a smarter model, not a new bug](https://daily.dev/posts/openai-was-breached-by-a-smarter-model-not-a-new-bug-bgdul0oun) · Kilo Blog · 0 upvotes · 0 comments

---

Tags: [#security](https://daily.dev/tags/security), [#cyber](https://daily.dev/tags/cyber), [#ai-agents](https://daily.dev/tags/ai-agents), [#data-breach](https://daily.dev/tags/data-breach), [#cisco](https://daily.dev/tags/cisco)

[View this post on daily.dev](https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl","headline":"AI models did the hacking at OpenAI, Cisco ships another perfect-10 CVE","text":"The OpenAI-via-Discourse story got a full technical writeup today, and the detail that matters is the exploit timeline: Claude Opus 5 wrote a working exploit within a day of release, GPT-5.6 Sol did it even better two days later. Cisco patched yet another maximum-severity ISE flaw under active exploitation. A zero-click plugin verification bug hit every major AI coding agent at once, and Gyazo confirmed 23.6 million records stolen from a server compromise.","url":"https://daily.dev/posts/ai-models-did-the-hacking-at-openai-cisco-ships-another-perfect-10-cve-6jigcpgdl","datePublished":"2026-09-19T04:18:35.527Z","dateModified":"2026-09-19T04:18:57.133Z","author":{"@type":"Organization","name":"Security Digest","logo":"https://media.daily.dev/image/upload/s--m4ZKB_C0--/f_auto,q_auto/v1779959612/logos/security_digest","url":"https://daily.dev/sources/security_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/security_digest","name":"Security Digest"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Security Digest","item":"https://daily.dev/sources/security_digest"},{"@type":"ListItem","position":3,"name":"AI models did the hacking at OpenAI, Cisco ships another perfect-10 CVE"}]}
```

