US Media: OpenAI Pitches New Cybersecurity Product, Briefs US Government and Five Eyes Alliance
OpenAI is pitching its new 'GPT-5.4-Cyber' security model to the US government and Five Eyes allies, adopting a dual-track strategy to bolster cyber defense and identify legacy system vulnerabilities.
📋 Article Processing Timeline
- 📰 Published: April 22, 2026 at 21:32
- 🔍 Collected: April 22, 2026 at 22:02 (29 min after Published)
- 🤖 AI Analyzed: April 23, 2026 at 14:32 (16h 30m after Collected)
(Central News Agency, New York, 22nd, Comprehensive Foreign Report) US news outlet Axios has learned that over the past week, artificial intelligence (AI) software giant OpenAI has continued to brief US federal agencies, state governments, and Five Eyes allies on the capabilities of its new cybersecurity product.
The report noted that OpenAI held an event in Washington, DC yesterday, demonstrating the functions of its next-generation GPT-5.4-Cyber model to approximately 50 federal government cybersecurity personnel. The model was launched last week under a tiered access program.
Sources revealed that to join OpenAI's "Trusted Access for Cyber" program, government applicants must go through the same vetting process as commercial clients.
Attendees at yesterday's event included officials from various government departments and national security agencies, most of whom are responsible for daily cybersecurity tasks.
Shortly before OpenAI launched its new cybersecurity model, AI company Anthropic also released Mythos Preview; both companies are working with government agencies to determine who gets access.
Due to cybersecurity risks, Anthropic has not publicly released Mythos, providing it only to about 40 enterprises and organizations, at least two of which belong to the federal government.
OpenAI has adopted a dual-track strategy, launching a version with stronger safety guardrails for broader use, while offering another version with looser usage restrictions to cybersecurity defenders through its "Trusted Access for Cyber" program.
OpenAI's President of Global Affairs Chris Lehane stated at yesterday's event that this approach will allow more companies, such as local water utilities, to access advanced AI tools.
OpenAI is also working with state governments to help them gain access to GPT-5.4-Cyber.
Notably, Axios learned that OpenAI began briefing Five Eyes alliance members this week to help them complete vetting and obtain model access. Members of this intelligence-sharing alliance include the US, Australia, Canada, New Zealand, and the UK.
Axios pointed out that most enterprises that have already obtained access to OpenAI's model and Anthropic's Mythos Preview are currently using these tools to find potential security vulnerabilities within their own internal systems.
In addition, many government agencies face difficulties in maintaining or patching legacy systems, and advanced AI tools are expected to help them discover critical vulnerabilities faster. (Compiler: Yang Chao-yen) 1150422
The report noted that OpenAI held an event in Washington, DC yesterday, demonstrating the functions of its next-generation GPT-5.4-Cyber model to approximately 50 federal government cybersecurity personnel. The model was launched last week under a tiered access program.
Sources revealed that to join OpenAI's "Trusted Access for Cyber" program, government applicants must go through the same vetting process as commercial clients.
Attendees at yesterday's event included officials from various government departments and national security agencies, most of whom are responsible for daily cybersecurity tasks.
Shortly before OpenAI launched its new cybersecurity model, AI company Anthropic also released Mythos Preview; both companies are working with government agencies to determine who gets access.
Due to cybersecurity risks, Anthropic has not publicly released Mythos, providing it only to about 40 enterprises and organizations, at least two of which belong to the federal government.
OpenAI has adopted a dual-track strategy, launching a version with stronger safety guardrails for broader use, while offering another version with looser usage restrictions to cybersecurity defenders through its "Trusted Access for Cyber" program.
OpenAI's President of Global Affairs Chris Lehane stated at yesterday's event that this approach will allow more companies, such as local water utilities, to access advanced AI tools.
OpenAI is also working with state governments to help them gain access to GPT-5.4-Cyber.
Notably, Axios learned that OpenAI began briefing Five Eyes alliance members this week to help them complete vetting and obtain model access. Members of this intelligence-sharing alliance include the US, Australia, Canada, New Zealand, and the UK.
Axios pointed out that most enterprises that have already obtained access to OpenAI's model and Anthropic's Mythos Preview are currently using these tools to find potential security vulnerabilities within their own internal systems.
In addition, many government agencies face difficulties in maintaining or patching legacy systems, and advanced AI tools are expected to help them discover critical vulnerabilities faster. (Compiler: Yang Chao-yen) 1150422