Breaking

OpenAI Agent Pipeline Produces 948 Working Linux Root Exploits in 30-Hour Run Costing $1.2 Million

November 4, 2026 at 23:32 EST

Security
AI Agents

An AI agent pipeline built by OpenAI generated 948 working Linux exploits capable of escalating a local user account to root privileges, according to figures reported by The Hacker News.

Reported figures · OpenAI-built AI agent pipeline

An AI pipeline generated 948 working root-escalation exploits in one 30-hour run

The run reportedly reproduced 1,724 vulnerabilities and built 291 information-leak proof-of-concept exploits, at a reported API cost of $1.2 million. The level of human oversight is not described.

948

Working Linux exploits for local-to-root escalation

1,724

Vulnerabilities reproduced

291

Proof-of-concept exploits for information-leak flaws

$1.2M

Reported total API cost over 30 hours

Output of one run, drawn to scale

Each block is roughly 1/10 of the column's count. All three values are counts of items.

1,724

Vulnerabilities reproduced

948

Working root exploits

291

Info-leak PoCs

The 948 exploits are roughly 55% of the reproduced vulnerabilities, a ratio the reporting does not explain in detail.

The pipeline, end to end

Discovery

Vulnerability data is identified

→

Reproduction

1,724 flaws re-created in test environments

→

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.