Endor Labs has introduced the Agentic Code Security Benchmark and the Agent Security League leaderboard to evaluate the security and functionality of AI-generated code, extending the SusVibes framework from Carnegie Mellon University researchers.
OpenAI has launched GPT-5.4-Cyber, a fine-tuned version of its GPT-5.4 model designed for defensive cybersecurity. The model is being released to vetted security vendors and participants of the Trusted Access for Cyber program, expanding access to advanced cyber defense tools.
A new survey from the Association for Financial Professionals shows that 76% of U.S. organizations experienced payments fraud in 2025, while only 17% are using AI to combat it. The report highlights ongoing vulnerabilities in checks and business email compromise.
Hadrius announced a new AI-native compliance infrastructure designed to embed intelligence directly into compliance systems, replacing traditional layered AI models. The system integrates data across communications, employee records, supervision structures, and policies to deliver real-time, AI-driven compliance actions.
Unisys has announced a partnership with Worldsys to integrate advanced anti-money laundering capabilities into its core banking platform, enhancing compliance and fraud prevention for financial institutions.
Electrosoft Services has been selected as one of four awardees on the Cybersecurity and Infrastructure Security Agency’s $500 million Digital Transformation Support Services agreement, enabling the company to support CISA’s IT modernization and cybersecurity initiatives over five years.
TrustCloud has released a native application for ServiceNow that provides continuous control monitoring for enterprise GRC and IRM users, syncing validated control signals directly into ServiceNow's IRM, SecOps, and CMDB modules.
SpartanX has announced a fully autonomous AI-powered red teaming platform that uses over 500 coordinated AI agents to conduct continuous, exploit-validated security testing across multiple attack surfaces.
Orca Security's 2026 State of Application Security Report reveals that 41.88% of production environments have leaked AI or ML credentials, with widespread vulnerabilities including unpatched Log4Shell exposures and weak CI/CD governance.
Anthropic has announced Project Glasswing, a cybersecurity collaboration with major technology and financial institutions to use its Claude Mythos AI model for detecting and fixing software vulnerabilities across critical systems worldwide.
ESET previewed new AI security tools at RSAC 2026 that scan chatbot prompts and responses to prevent data leaks and malicious content, with expanded capabilities for endpoint protection and AI supply chain defense.
Microsoft announced a $10 billion investment in Japan from 2026 to 2029, focusing on AI infrastructure, cybersecurity partnerships, and training one million workers. The initiative includes collaborations with Sakura Internet and SoftBank to expand domestic AI compute capacity and support Japan’s data sovereignty goals.
Ariel Simon of Token Security will present research at RSAC 2026 detailing a remote code execution flaw in Microsoft’s Azure Model Context Protocol server that could allow attackers to compromise cloud environments.
Netwrix has introduced new capabilities to its 1Secure platform that provide visibility and control over how AI agents, including Microsoft Copilot, access sensitive data across hybrid environments.
Onyx Security has launched with $40 million in funding from Conviction and Cyberstarts to develop a secure AI control plane that helps enterprises manage and govern AI agents.
Quantro Security has emerged from stealth to launch VM.Analyst, an autonomous AI agent designed to modernize enterprise cybersecurity by automating risk analysis and remediation.
CrowdStrike and Perplexity have partnered to bring the Falcon platform’s real-time threat detection and governance capabilities to the Comet Enterprise AI browser, offering enterprises enhanced data protection and visibility within AI-driven workflows.
The University of Tulsa has secured $12.6 million in federal appropriations to advance research in AI, cybersecurity, materials, and energy systems, expanding its pilot-scale facilities and industry partnerships.
Cato Networks has initiated the FedRAMP High Authorization process for its SASE Platform in collaboration with Coalfire, aiming to expand its presence in the U.S. federal market and meet growing security requirements.
SecureIQLab has released version 5.0 of its Cloud WAAP CyberRisk Validation Methodology, the first to test AI-powered defenses using AI-powered attacks. The new framework expands testing to cover AI-assisted bots, API gateways, and LLM-integrated application stacks.
Weekly newsletter about AI in Cybersecurity.