AI News

GPT-6 Cyber Finds and Fixes Nearly 950 Linux Kernel Vulnerabilities in a Single Scan

DeepLearning.AI · Letter from Andrew NgUpdated Nov 13 23:55

AI summary

OpenAI has launched GPT-6 Cyber, a security fine-tuned version of GPT-6 Astra. In a single scan of the Linux kernel in mid-October, it addressed 948 vulnerabilities that could allow an ordinary user to gain full control of a machine, at an average cost of about $1,280 per vulnerability. The model supports up to 1,050,000 input tokens and 128,000 output tokens, with API pricing at $20/$2/$100 per million input/cached/output tokens (when input does not exceed 272,000 tokens). It is currently available only to selected customers in OpenAI's Daybreak Red advanced security program.

Body

Only a summary is authorized here. Continue at the validated source link.

Source

DeepLearning.AI · Letter from Andrew Ng