# Latent Space reports AI security breach wave and dozens of model releases

Digest AI · Business & Funding · published 2026-10-23T10:09:29Z

Canonical: https://digestai.news/story/latent-space-reports-ai-security-breach-wave-and-dozens-of-model-relea

## Summary

A rapid 48‑hour wave of AI‑accelerated vulnerability disclosures overwhelmed defenders, with critical flaws reported in PHP (PHP2Shell RCE), GitLab (CVSS 10), Node.js, and NGINX. Supply‑chain incidents included a password‑stealing LiteLLM 1.107.0 release, a breach of 321,810 Zeabur accounts, and weak CA keys at eMudhra that Claude allegedly cracked. The CVE count rose from about 4 k to about 15 k since April, and a China‑aligned group used the ARTEX pentest framework against over 2,250 Microsoft 365 tenants.

OpenAI and Anthropic announced large‑scale security tooling, with Anthropic’s Frontier Red Team checking 180 M lines of code across 123 libraries, generating 239 entry points and shrinking MiniMax M3.1 Flash’s checkpoint by 44.6% to 250.1 GB. New open‑weight models debuted, including MiniMax M3.1 Flash (428 B total, 1 M token context), Carbon‑X (14 B DNA‑plus‑text, Apache 2.0), OCRia‑1 (4 B VLM), Flint (40 B open model), and Canva Craft (image model). Funding news featured OpenAI’s reported near‑$30 B raise at a $1.25 T valuation led by UAE investors, Warp Sciences’ $215 M Series A, and Datadog’s $4 B offer for Upwind Security.

## Key points

- 48‑hour wave of AI‑accelerated vulnerability disclosures hit 12 subreddits, 544 tweets, with critical flaws in PHP, GitLab, Node.js, NGINX.
- MiniMax M3.1 Flash (428 B total, 23 B active) offers 1 M token context, 73.7% DeepSWE score, 44.6% checkpoint shrink to 250.1 GB.
- OpenAI reportedly near a $30 B raise at a $1.25 T valuation, led by UAE investors; Datadog offered ~$4 B for Upwind Security.

## Why it matters

The surge of AI‑driven attacks shows growing security risks as models become more capable, while rapid model releases and massive funding reshape competition and accelerate deployment across industries.

## Sources

1. [AINews Black October](https://latent.space/p/ainews-black-october) (Latent Space, 2026-10-23)

Part of the developing story: [AI Model Race Accelerates with New Releases](https://digestai.news/thread/typesafe-ai-releases-jev-1-14-with-images-box-type-and-262k-token-context) (3 stories)

## Cite

Digest AI, "Latent Space reports AI security breach wave and dozens of model releases", 23 October 2026, https://digestai.news/story/latent-space-reports-ai-security-breach-wave-and-dozens-of-model-relea

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/latent-space-reports-ai-security-breach-wave-and-dozens-of-model-relea.json