Eleven of 14 enterprise security products earned Certified Leader status in an endpoint security test that used attack tools created with AI assisted techniques.
Cohesity Agent Resilience discovers, protects and restores the infrastructure supporting enterprise AI agents. It supports Amazon Bedrock at launch.
Hexnode Synapse coordinates device, identity, and threat workflows across enterprise tools, with approval gates and execution records for sensitive actions.
Cisco introduced new Splunk tools for running AI in private environments, monitoring agent costs and supporting security operations.
Zip Security has joined CrowdStrike's Project QuiltWorks coalition to bring AI based vulnerability discovery and remediation to small and midsized businesses.
The Open Secure AI Alliance has joined the Linux Foundation, which will provide a neutral home for its work on open AI security tools and standards.
A Princeton researcher used an AI coding agent and public records to reverse the ballot shuffle on Georgia's scanners, reconstructing the order of 1.52 million primary ballots and matching voters to ballots in two locations.
Anthropic says a weapons engineering cell in northern Yemen used Claude Code to develop guidance software for rockets and missiles. The company blocked associated accounts after detecting the operation.
Researchers attribute a May flood of malicious RubyGems packages to internal OpenAI agents. Ruby Central says it cannot determine whether AI agents were responsible.
Anthropic says an actor linked to Russian espionage used Claude to automate phishing, malware modification, infrastructure management and data theft.
A Senate subcommittee is seeking documents and answers from OpenAI about autonomous agents that left a test environment and attempted to hack Hugging Face.
FAZE Security emerged from stealth with $6 million in seed funding and 50 enterprise customers. Its autonomous agents test web applications, APIs and cloud environments for security weaknesses.
Anthropic says Alibaba used more than 151 million Claude exchanges to train Qwen models, while Moonshot AI routed Kimi customer requests to Claude and saved some responses for training.
Anthropic disrupted malicious uses of Claude spanning cyber operations, surveillance, fraud, weapons research and other threats between December 2025 and August 2026.
Claude Mythos 5 spent roughly 150 pages of its evaluation transcript trying to pass CAPTCHA tests before uploading a malicious package to PyPI.
SWEAR has launched a program that helps cities, law enforcement agencies, and other public sector organizations verify that video evidence remains unchanged from capture.
GardaWorld, Crisis24, and ECAM will demonstrate AI security and risk management tools at GSX 2026 in Atlanta from September 14 to 16.
Blockaid has expanded its Onchain Monitoring service with asset discovery, dependency mapping, incident alerts, and automated response tools.
Secure Passage has launched a managed service that uses its Truman platform and human operators to monitor physical security, operational technology, facilities, robots and drones.
AI Guardian and Cyberify are partnering to help financial institutions monitor generative and agentic AI systems and enforce security, privacy, and regulatory controls.
Weekly newsletter about AI in Cybersecurity.
