Check Point ruster kunder med OpenAIs mest avancerede AI-modeller til cyberforsvar

On June 22, 2026, Check Point announced it is embedding OpenAI's frontier AI models into its customer-facing security products as a participant in the OpenAI Daybreak Cyber Partner Program, moving the partnership beyond internal defensive use to direct customer protection . The move places Check Point among a select group of vetted security vendors and signals an escalation in the AI-powered cybersecurity arms race, as both OpenAI and Anthropic race to deploy frontier AI models for defensive security operations.

Selectivity and Phased Rollout

The Daybreak Cyber Partner Program is described as open to only a select group of vetted security vendors . Check Point had already been approved for OpenAI's "Trusted Access for Cyber" (TAC) program on June 11, 2026, gaining early access to GPT-5.5 for internal defensive operations before graduating to the broader Daybreak Partner Program for customer-facing integration . This phased approach allowed Check Point to test and validate the frontier models in controlled defensive workflows before embedding them into products used by its customers.

Shift from Internal Use to Customer-Facing Defense

Check Point's initial TAC membership let it use GPT-5.5 internally for high-stakes defensive security operations like threat analysis, incident investigation, and real-time detection development . The June 22 announcement marks a direct escalation: those same frontier AI capabilities are now being embedded into the security products and workflows that Check Point's 100,000+ customers use every day . As Check Point's press release states, "It marks a meaningful shift, from using these models internally to embedding them directly inside the defenses that protect customers" .

Goal: Sharper Threat Prevention and Faster Remediation

The stated goal is to embed frontier AI reasoning directly into defensive workflows — improving threat prevention accuracy and accelerating the remediation lifecycle . OpenAI's Daybreak initiative as a whole is designed to move from vulnerability discovery to end-to-end patch automation at machine speed . This includes the full release of GPT-5.5-Cyber, an updated Codex Security plugin, and the open-source Patch the Planet initiative, all announced on June 22, 2026 .

Earlier Launch of Agentic Exposure Validation (AEV)

On May 28, 2026, Check Point launched Agentic Exposure Validation (AEV) as part of its Exposure Management portfolio . AEV uses AI agents that "reason like attackers" to determine whether exposures are actually exploitable, producing evidence and remediation guidance rather than relying on static severity scores . Yochai Corem, Check Point's GM of Exposure Management, stated: "The era of autonomous, AI-driven exploitation is here. Frontier AI models are attacking critical vulnerabilities at scale, without human steering."

Collaboration on Responsible AI Standards and Safety Controls

The Daybreak Cyber Partner Program is structured as a governed access framework — partners are vetted and given access under OpenAI's safety and responsibility protocols . The press materials frame the partnership as embedding "frontier-grade" engines inside controlled defensive environments, with safety controls built into the partner program model itself . GPT-5.5-Cyber is released only through a "limited release to trusted defenders," with verification and monitoring, while a milder GPT-5.5 with Trusted Access for Cyber is available for general defensive workflows . No separate, standalone agreement on responsible AI standards beyond the Daybreak program's governance structure is detailed in the available sources.

Competitive Context: Daybreak vs. Anthropic's Project Glasswing

OpenAI's Daybreak and Anthropic's Project Glasswing are two competing frontier-AI cybersecurity programs launched in 2026. Both are responses to the same trend — frontier models becoming capable of autonomous vulnerability discovery and exploitation — but they take different approaches:

Aspect	OpenAI Daybreak	Anthropic Project Glasswing
Launch date	May 11, 2026 (initial); expanded June 22, 2026	April 2026 (initial ~50 partners); expanded to ~150-200 orgs by June 2, 2026
Frontier model	GPT-5.5 / GPT-5.5-Cyber	Claude Mythos Preview
Focus	Vulnerability discovery + end-to-end patch automation; partner program for security vendors	Vulnerability hunting; critical-infrastructure organizations test defenses using the model
Partner model	Daybreak Cyber Partner Program — vetted cybersecurity vendors embed model into customer products	Direct access for ~150-200 critical infrastructure orgs, including Microsoft, Nvidia, Google
Scale (June 2026)	Multiple new partners announced simultaneously (Check Point, Darktrace, Sophos, Proofpoint, SpecterOps, IBM, TrendAI)	Expanded to 150-200 organizations across 15+ countries; partners reported 10,000+ high/critical findings

Daybreak leans on a vendor-partner distribution model, where vetted cybersecurity companies embed the models into their own products to reach enterprise customers. Glasswing, in contrast, gives direct model access to critical infrastructure operators and major tech partners, who use the Claude Mythos Preview model to scan their own codebases for vulnerabilities . As of June 2026, Glasswing partners had collectively reported over 10,000 high- and critical-severity security flaws in under two months of operation .

The Bigger Picture: Why This Matters

The Check Point announcement is part of a broader industry shift. On June 22, 2026, OpenAI simultaneously announced the expansion of its Daybreak program with multiple new partners (Check Point, Darktrace, Sophos, Proofpoint, SpecterOps, IBM, TrendAI), the full release of GPT-5.5-Cyber, an updated Codex Security plugin, and the open-source Patch the Planet initiative . The Patch the Planet initiative, founded with security firm Trail of Bits and in collaboration with HackerOne, funds security researchers to work with open-source project maintainers to patch critical vulnerabilities . This flurry of activity signals that the era of AI-powered cybersecurity — both offensive and defensive — has arrived at machine speed.