OpenAI briefs governments and Five Eyes on GPT-5.4-Cyber model amid AI security race

Exclusive: OpenAI briefs feds and Five Eyes on new cyber product

Why it matters: Companies and agencies are clamoring to get their hands on the latest AI tools, whose advanced cybersecurity capabilities promise big gains for defenders and frightening advances for malicious hackers. Driving the news: OpenAI held an event in D.C. on Tuesday for approximately 50 cyber defense practitioners across the federal government to demo the capabilities of its new GPT-5.4-Cyber model, which it rolled out under a tiered access program last week. * Government applicants are going through the same vetting process as commercial customers who wish to join its Trusted Access for Cyber program, a source familiar said. * Attendees included a range of officials from across the government and national security agencies, most of whom oversee day-to-day cyber tasks. Split screen: OpenAI rolled out its new cyber model hot on the heels of Anthropic's Mythos Preview, and both companies are currently working with government agencies to determine who will have access. * Anthropic withheld a public release of Mythos, citing its cyber risks, and offered it only to around 40 companies and organizations, including at least two in the federal government. * OpenAI is pursuing a dual-track approach of making one version of its model more widely available with strong safeguards in place, while releasing another more cyber-permissive version to defenders through the Trusted Access program. Zoom in: At Tuesday's event, OpenAI Chief Global Affairs Officer Chris Lehane said that approach would allow more companies, like local water utilities, to access advanced AI tools. * Sasha Baker, OpenAI's head of national security policy, told attendees that OpenAI hopes to partner with government departments to prioritize the most crucial use cases and build channels to share threat intelligence across sectors. * The company is also working with state governments to get them access to GPT-5.4-Cyber. The intrigue: OpenAI is starting briefings with Five Eyes members this week to get them vetted and signed up to access the model, Axios has learned. * In addition to the U.S., that intelligence-sharing partnership includes Australia, Canada, New Zealand and the U.K. Anthropic's own rollout within the U.S. government is complicated by the Pentagon's decision to label the company as a "supply chain risk" after a messy AI safeguards fight. * Still, Mythos is currently being tested by the NSA despite the designation, as Axios reported. Between the lines: Most companies that already have access to OpenAI's model and Anthropic's Mythos Preview are using the tools to find exploitable security flaws in their own internal systems.

OpenAI, Microsoft expand cybersecurity partnership amid rising AI threats

OpenAI and Microsoft are joining forces to combat cyber threats. They will deploy advanced artificial intelligence tools to enhance security. OpenAI's powerful AI models will be accessible to Microsoft. Microsoft will use its cybersecurity expertise to protect OpenAI's systems and customers. This collaboration aims to strengthen defenses against evolving cyber risks. OpenAI and Microsoft on Thursday announced an expanded cybersecurity partnership to deploy advanced artificial intelligence (AI) tools against cyber threats. In a post on social media platform X on Thursday, the company said OpenAI will provide Microsoft access to its most advanced cyber-capable AI models through its Trusted Access for Cyber programme. Microsoft, in turn, will apply its cybersecurity infrastructure and expertise including its Secure Future Initiative, to help protect OpenAI's systems, models and shared customers. The companies did not disclose financial terms or a timeline for the deal. Introduced in February this year, OpenAI's pilot "Trusted Access for Cyber," is an identity-based framework designed to ensure that its most powerful cybersecurity capabilities are primarily deployed by vetted users. OpenAI spent $10 million in API credits through its Cybersecurity Grant Programme to accelerate adoption among security teams, particularly those focused on identifying and fixing vulnerabilities, per a company blog post. Cybersecurity in focus The announcement comes as a response to rapid advances in AI. "AI models are becoming much more capable in cybersecurity, and that progress raises the bar for everyone," OpenAI said. The development follows rising cybersecurity issues due to Anthropic's Mythos model, which Microsoft itself plans to embed into its secure coding framework, as the company steps up its cybersecurity capabilities. Microsoft recently evaluated Mythos, using its own open-source benchmark for real-world ⁠detection engineering tasks, and the "results showed substantial improvements relative to prior models." OpenAI's push for cyber defense OpenAI, meanwhile, unveiled GPT-5.4-Cyber earlier this month, which is a variant of its latest flagship ⁠model fine-tuned specifically for defensive cybersecurity work. The ChatGPT-maker has been resilient in building cyber defense capabilities. OpenAI said its cybersecurity strategy centers on three principles: expanding access to legitimate defenders, deploying systems iteratively, and investing in ecosystem resilience. It also held ⁠an event in Washington for about 50 cyber defense practitioners across the federal government to demo the capabilities of its new GPT-5.4-Cyber model. The Sam Altman-led company has been ramping up efforts since December last year, saying it is "investing in strengthening models for defensive cybersecurity tasks". In a blog post published earlier this month, OpenAI said it has supported defenders since 2023 through its Cybersecurity Grant Programme and has expanded tooling such as Codex Security, which automatically scans codebases, validates vulnerabilities and proposes fixes. According to the company, the system has contributed to fixing more than 3,000 critical and high-severity vulnerabilities since its recent rollout. The firm also said it has reached more than 1,000 open-source projects with free security scanning tools.

OpenAI Briefs Governments On New Cyber AI Model Amid Anthropic Claude Rivalry

OpenAI met with 50 cyber defense practitioners across various federal agencies, state governments, and Five Eyes allies to unveil its new GPT-5.4-Cyber model. The artificial intelligence company held an event in D.C. on Tuesday to demo the capabilities of this new cyber model, which debuted as a tiered access program last week, Axios reported. The model will be deployed through a dual-track approach, with one version broadly accessible and equipped with strong safeguards, and a separate, more cyber-capable version reserved for defenders via the Trusted Access program. The pilot program was announced in February in an effort to "enhance baseline safeguards for all users while piloting trusted access for defensive acceleration." The model will also be available via an "intelligence-sharing partnership" that includes Australia, Canada, New Zealand, and the U.K., Axios noted. GPT-5.4-Cyber rivals Anthropic's Claude Mythos Preview, a model which is used to hunt and fix software flaws in an effort to "reshape" cybersecurity. Anthropic expanded access to Mythos to more than 40 additional organizations involved in critical software infrastructure, covering both proprietary and open-source code. This move comes as AI capabilities surrounding cybersecurity have reached a "tipping point." Government officials have recently raised concerns that artificial intelligence tools could be misused to disrupt critical infrastructure such as financial systems or power grids. Both OpenAI and Anthropic have engaged with government agencies (including defense and public-sector organizations) to deploy or evaluate AI systems in controlled settings, often with a focus on security, safety, and sensitive use cases. It was recently reported that the U.S. National Security Agency (NSA) is allegedly using Anthropic's restricted artificial intelligence model, Claude Mythos Preview, for cyber defense despite the Pentagon blacklisting the company by designating it a supply chain risk. Photo: Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

OpenAI Begins Briefing Governments on Cybersecurity Capabilities | PYMNTS.com

The AI startup held an event in Washington D.C. Tuesday (April 21) where it demonstrated its new GPT-5.4-Cyber model, Axios reported. The report, citing a source familiar with the matter, said attendees included officials from throughout the government and from various national security agencies, most of whom are in charge of day-to-day cyber tasks. OpenAI is also working with state governments to get them access to GPT-5.4-Cyber, the report said, and is beginning to brief the "Five Eyes," a multi-national intelligence-sharing alliance made up of the U.S., Canada, the U.K., Australia and New Zealand. The company is taking a dual-track approach, the report said, making one version of its model more widely available with robust safeguards, and another more permissive version for cyber defenders through its Trusted Access program. OpenAI Chief Global Affairs Officer Chris Lehane said during Tuesday's meeting that this tactic will let more companies, like local water utilities, access advanced AI tools. Sasha Baker, who heads the company's national security policy, told attendees that OpenAI hopes to work with government departments to prioritize the most crucial usages and share threat intelligence across sectors. This effort is happening soon after OpenAI rival Anthropic began previewing its Mythos AI model. That startup has held off on a wide release of Mythos, claiming it is too dangerous. Instead, Anthropic has offered the model to around 40 companies and organizations, including some within the government. The report came the same day that Anthropic said it was investigating reports that a small group of people had gained access to Mythos through a third-party vendor. The company told PYMNTS it had found no evidence this activity extended beyond that vendor. Meanwhile, PYMNTS wrote last week about an evaluation of Mythos by the U.K. Government's AI Security Institute (AISI). The chief takeaway from those findings, the report said, is not that AI can already carry out flawless cyberattacks. In fact, the AISI report noted that the success rate is limited. "But systems that can plan and execute multistage intrusions, even inconsistently, represent a baseline that will improve," PYMNTS added. "More compute, better orchestration, and tighter integration with external tools will incrementally close the gap between partial and reliable capability."