Safety and ethics
Redwood says OpenAI agents disable safety monitoring
Redwood Research says several OpenAI agents disabled deviation monitoring for the models they were running, and in some cases other agents were launched without…
CrowdStrike launches AI-powered red teaming for agentic attack paths
CrowdStrike has introduced Agentic Adversary Emulation, an AI-powered red teaming capability designed to show how attackers could chain weaknesses across systems and…
CrowdStrike Adds AI Agent Insider Threat Detection to Falcon
CrowdStrike is extending its agentic SOC with capabilities designed to detect compromised AI agents, investigate their behavior and connect risks across agents, users…
CrowdStrike Adds Cloud Security Tools to Protect AI Infrastructure
CrowdStrike has introduced generally available Falcon Cloud Security innovations designed to protect software supply chains, AI workloads and cloud infrastructure from…
Binance hack and AI-discovered vulnerabilities: what is really changing?
More than $6 billion worth of digital assets were stolen from the Binance cryptocurrency exchange, while artificial intelligence was used to uncover severe software…
Record $13 Million Cybersecurity Settlement Reached
ILC Dover and ILC Astrospace will pay the United States $13 million to resolve a case in which the companies were accused of making false claims about cybersecurity…
Aurora Driver puts the safety of first responders and police first
Aurora says its self-driving systems have been designed from the outset to handle the presence of first responders, police officers and other public-service workers…
Darkforce’s cybersecurity platform receives Awardable designation
Darkforce’s cyber and artificial intelligence security platform has received Awardable status on the US Chief Digital and Artificial Intelligence Office’s Tradewinds…
Anaconda's Enkrypt AI featured in two Gartner reports
According to Anaconda, Enkrypt AI is mentioned as a Sample Vendor for AI security testing in two 2026 Gartner Hype Cycle reports. The documents examine AI-driven…
Safety and ethicsCogent’s VR-1.1 model fixed 57.9% of vulnerabilities
Cogent Security has released VR-1.1, a cybersecurity reasoning model that produced working fixes for 57.9% of the public LiveCVEBench test, according to the company’s…
Bold Security makes the 2027 Fortune Cyber 60 list
Bold Security, which develops an on-device AI-powered data loss prevention solution, has been named to the 2027 Fortune Cyber 60 list. The company appears in the…
A single HTTP request can completely take over the Reachy Mini AI robot
Five vulnerabilities in the Reachy Mini AI robot could allow an attacker on the same network to take control of the device without authentication. JFrog says the attack…
AFRL brings Stream Engine deepfake analyzer to test-ready status
Stream Engine has delivered its semantic-based deepfake detection analyzer to the US Air Force Research Laboratory. By June 2026, the system had entered the Air Force…
Hidden advertising found in package targeting OpenAI Codex
Check Point Research identified a method that persistently alters an AI agent’s trusted instructions so that advertising appears in later responses. The Haiou OpenAI…
Cognition says companies can protect themselves with basic security measures
Recent major security incidents show that a compromised developer machine, a stolen credential or an unpatched system can cause serious damage on its own. Cognition has…
CrowdStrike launches AI Kill Switch to stop rogue AI agents in Falcon Guardian
CrowdStrike has introduced AI Kill Switch, a control in Falcon Guardian that lets security teams stop a misbehaving AI agent by blocking its access to the AI models…
CrowdStrike adds AI kill switch to Falcon Guardian to stop rogue agents
CrowdStrike has launched an AI kill switch for its Falcon Guardian product, letting security teams cut off AI model access at AI gateways when agents become dangerous.…
CrowdStrike brings Falcon sensor inside Anthropic's Claude Cowork VM
CrowdStrike is bringing Falcon Guardian protection natively into Anthropic's Claude Cowork, running the Falcon sensor directly inside the virtual machine where the agent…
E2B Workload Identity lets backends verify which sandbox is calling
E2B has launched Workload Identity, a feature that attaches a signed JSON Web Token (JWT) to outgoing sandbox requests so a backend can verify which project, sandbox and…
Resemble AI launches native real-time deepfake detection in Microsoft Teams
Resemble AI has launched a native real-time deepfake detection integration for Microsoft Teams. It scans audio and video during meetings for synthetic media, identity…
Vercel Security Review enters public beta for Pro and Enterprise teams
Vercel Security Review is now in public beta for Pro and Enterprise teams. Once enabled, every push to a connected GitHub repository is automatically reviewed for…