The Information Machine
Concluded·following since 10 Aug 2026·Day 6·11 sources·updated 18 Aug 2026

OpenAI's GPT-5.6-Cyber and Daybreak program

The gist

OpenAI's GPT-5.6-Cyber, a 95%-benchmark cybersecurity model, reaches enterprise clients through Daybreak partners and AWS

A purpose-trained model completing 95% of advanced cybersecurity tasks, with documented zero-day discoveries, represents a substantial capability step in AI-assisted offensive and defensive security work. The program's design, routing access through enterprise partners rather than end clients, is OpenAI's explicit approach to managing the dual-use risks of a model built to approve exploit-development requests.

The full picture

GPT-5.6-Cyber is OpenAI's purpose-trained model for advanced cybersecurity tasks including exploit development, vulnerability research, and penetration testing, available through the Daybreak partner program and Amazon Bedrock. Daybreak has two tiers: Daybreak Blue, giving approved partners access to GPT-5.6 Sol with defensive safeguards removed, and Daybreak Red, providing GPT-5.6-Cyber for higher-risk offensive and research work. Access goes through a roster of approved enterprise security partners rather than direct signup; named partners include Accenture, IBM, Capgemini, EY, KPMG, PwC, Palo Alto Networks, CrowdStrike, Cisco, Cloudflare, NCC Group, SpecterOps, Sophos, Akamai, Fortinet, and Cognizant. End clients of those partners receive only the findings from security work, not the underlying model.

On OpenAI's internal advanced cybersecurity benchmark, GPT-5.6-Cyber scores 95.0%, compared to 57.3% for the prior GPT-5.5-Cyber, 2.0% for GPT-5.6 Sol via Daybreak Blue, and 1.5% for GPT-5.6 Sol under standard safeguards. OpenAI used the model to discover CVE-2026-15903, described as two chained zero-days in Chrome's V8 engine, over 400 kernel privilege-escalation vulnerabilities, and critical flaws in a popular database and mobile OS. Pricing is $12.50 per million input tokens and $75 per million output tokens for GPT-5.6-Cyber, versus $5/$30 for GPT-5.6 Sol. The model is rated High under OpenAI's Preparedness Framework, one level below Critical, and mandatory hardware security keys for all Daybreak accounts take effect September 1, 2026.

Daybreak models are accessible via Amazon Bedrock using the bedrock-mantle endpoint, with Daybreak Access approval required. A subsequent report characterized the AWS path as removing the prior requirement for personal vetting by OpenAI, with access described as provisioned "through your company's existing Amazon account, built for speed, not scrutiny".

How it developed
18 August 2026

A report published August 17 characterized Amazon Bedrock access to Daybreak models as removing OpenAI's prior personal vetting requirement, describing provisioning as happening "through your company's existing Amazon account, built for speed, not scrutiny." That characterization contradicts OpenAI's own August 11 announcement, which stated that Daybreak Access approval remains required for the bedrock-mantle endpoint.

GPT-5.6-Cyber, a purpose-trained cybersecurity model scoring 95% on OpenAI's internal benchmark, became available through Amazon Bedrock on August 11 as part of a restructured two-tier Daybreak partner program.

First citedThe Neuron
17 August 2026

Report characterized the AWS path as removing personal vetting, provisioning through existing Amazon accounts

Reporting on August 16 confirmed pricing and governance details for the security model OpenAI launched August 10: GPT-5.6-Cyber is priced at $12.50 per million input tokens and $75 per million output, against $5 and $30 for GPT-5.6 Sol, and the Daybreak partner program moved from three loosely defined prior tiers to two clearly named ones, with hardware security keys mandatory for all accounts from September 1, 2026. One analysis also flagged that two aspects of the benchmark chart showing GPT-5.6-Cyber completing 95% of advanced cybersecurity tasks warrant more scrutiny than the headline figure, without elaborating further.

14 August 2026

A commentary piece published August 13 described OpenAI as occupying a position where it could be both a source of and a defense against cybersecurity crises, following the August 10 release of GPT-5.6-Cyber, a model purpose-trained for exploit development, and the expansion of the Daybreak Cyber Partner Program to security and consulting firms including Palo Alto Networks, CrowdStrike, and IBM.

The Neuron Daily also reported that GPT-5.6-Cyber's pricing has not been made public.

13 August 2026

Commentary noted the irony that OpenAI could be both the cause of and solution to cybersecurity crises

11 August 2026

Daybreak models made available through Amazon Bedrock via the bedrock-mantle endpoint

10 August 2026

Sam Altman encouraged users to use OpenAI models to defend their systems

Sources
6 more sources
The daily email

Want this in your inbox?

I send a short email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free