OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs

OpenAINov 4, 6:50 PM PT

From OpenAI: OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs — Security research that once required a separate, sustained investigation for each finding can increasingly be run as a systematic…

Read the original at openai.com ↗Techmeme ↗

Source: OpenAI (openai.com)
Published: Nov 4, 6:50 PM PT
First seen here: Nov 4, 7:53 PM PT
Edition: Wednesday, November 4, 2026, ranked 7
Status: The company itself announced or confirmed this.
Topics: AI, Security, Open source, Openai

News

Partnering with Hacktron to find vulnerabilities in OpenAI’s systems (opens openai.com)

OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services, and strengthen defenses with frontier AI.

OpenAI12:00 AM PTNov 7, 2026 Details

Launch

GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents (opens openai.com)

OpenAI's cybersecurity model, a version of GPT-6 Astra, finds exploitable vulnerabilities and develops fixes while staying within authorized scope.

OpenAI1:00 AM PTNov 5, 2026 Details

DiscussionUnverified claim

OpenAI seems to be silently re-enabling model training on Codex chats (opens news.ycombinator.com)

A poster says the data-sharing setting their team had turned off in Codex was switched back on without anyone opting in.

Hacker News6:53 AM PTNov 3, 2026 Details

Launch

Anthropic releases FlashProve, an open-source prototype that cryptographically verifies an AI response came from a given model, not one that a router swapped in (opens anthropic.com)

The prototype would let anyone check that a response really came from the model it claims to be from.

Anthropic1:10 PM PTNov 3, 2026 Details

News

Researchers are paid $5K for reporting a bug to OpenAI that would let attackers run code via a ChatGPT share link (The Hacker News) (opens thehackernews.com)

Researchers are paid $5K for reporting a bug to OpenAI that would let attackers run code in logged-in sessions of all who opened the hacker's ChatGPT share link — A malicious ChatGPT share link could run an attacker's code in the logged-in session of anyone who opened it …

thehackernews.com12:01 PM PTOct 31, 2026 Details

News

Making it easier for websites to express AI training preferences (opens openai.com)

Google, Microsoft, OpenAI and Anthropic will honour one shared robots.txt rule that keeps a site's content out of generative AI training.

OpenAI2:30 AM PTOct 31, 2026 Details

OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs · Keeping Up with Silicon Valley

OpenAI details a pipeline that used an upcoming cyber model to validate and patch 948 local privilege escalation vulnerabilities in Linux for $1.2M in API costs

Related stories

Partnering with Hacktron to find vulnerabilities in OpenAI’s systems (opens openai.com)

GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents (opens openai.com)

OpenAI seems to be silently re-enabling model training on Codex chats (opens news.ycombinator.com)

Anthropic releases FlashProve, an open-source prototype that cryptographically verifies an AI response came from a given model, not one that a router swapped in (opens anthropic.com)

Researchers are paid $5K for reporting a bug to OpenAI that would let attackers run code via a ChatGPT share link (The Hacker News) (opens thehackernews.com)

Making it easier for websites to express AI training preferences (opens openai.com)