‘Sovereign AI agents’ pose threat to human civilisation, report warns
The Independent — Congress should hold those who release sovereign AI agents liable for the harms that follow, experts say
Safety & Security
Artificial intelligence safety and security encompasses the technical and policy measures used to mitigate model risks, vulnerabilities, abuse and failures. Truncated covers research developments, threat analyses, defense mechanisms and policy discussions related to reliable deployment.
OpenAI employees warned about operational security lapses for years, sources say
The Japan Times — Workers said OpenAI’s controls were so lax that outsiders could obtain sensitive information about the AI lab’s technology with relative ease.
AI ‘godfather’ Geoffrey Hinton says it might turn out humans would have been better off without AI
the Guardian — Nobel winner who devised concept of neural network warns AI could be near to triggering major incident such as power or banking shutdown
Siri AI will not get real chatbots until Apple can secure them
AppleInsider — John Ternus has said that Apple wants Siri AI to offer agentic AI tools, but not until it can be sure users truly understand what the risks are.
OpenAI says its new cybersecurity model is better at not hacking other companies
The Verge — GPT-6 Cyber is a big step up from OpenAI’s last cybersecurity model.
Cyberhaven launches Agentic Data Security to track and control AI data use
Help Net Security — Cyberhaven Agentic Data Security allows enterprises to say yes to AI agents without losing control of their data.
Arkose Labs expands Agent Trust Manager to govern agent sessions
Help Net Security — Arkose Labs expands Agent Trust Manager to classify agent traffic, enforce policies and track agents across networks and accounts.
An AI Fraud Kingpin’s Confessions
404 Media — An AI fraud kingpins rise and fall; voters rejected data centers; and a new fear unlocked.
PortSwigger’s Burp AT lets enterprises delegate pentests with defined limits
Help Net Security — PortSwigger Burp AT brings agentic pentesting to enterprises, with defined test scopes, enforced permissions and expert oversight.
Sysdig AI Defense puts runtime guardrails on AI agents
Help Net Security — Sysdig AI Defense monitors agents at runtime, enforces access policies and blocks risky actions across endpoint and cloud systems.
Agentic AI Is Shipping Code. Who Owns The Risk?
Forbes — Any approach to governing agentic AI has to start with knowing where it is being used.
Zero Trust Agent Systems that Pass Audits and Still Ship
InfoQ — Advait Patel shares a playbook for securing AI agent systems in production. He discusses moving defense boundaries from the prompt to the tool layer, enforcing scoped workload identities, and mitigating prompt injection via tool outputs. He explains how to structure audit-ready observability and implement kill switches to protect critical data while keeping deployment fast and compliant.
Android Auto’s endless calling loop is driving Pixel and Galaxy owners crazy
Android Authority — Pixel and Galaxy owners say Android Auto keeps asking who to call or text, even after recognizing their contacts.
One Hacker Uses 12 AI Agents to Breach Taiwan's Emergency Services and Steal Sensitive Data
Cyber Security News — A Taiwan-based operator used 12 AI agents to attack hundreds of organizations across 11 countries, stealing sensitive data.
Tokenmaxxing crackdowns and AI cybersecurity risks
IT Pro — As companies cut down on ‘tokenmaxxing’, what are the risks posed by the practice and how will crackdowns impact cybersecurity?
AI agent detection signals that still work when agents look human
Help Net Security — Detecting AI agent traffic is hard when undeclared and disguised sessions behave like people and slip past binary bot tools.
India Warns AI Hackers Are Targeting Crypto Exchanges
Cointelegraph — India’s CERT-In warns that AI-assisted hackers are targeting crypto exchanges and financial firms, stealing wallet keys and draining funds in an ongoing campaign.
A thousand fewer paths to root
Real application testing keeps AI-driven security patches safe
SiliconANGLE — Real application testing lets customers replay live workloads to check that large AI-driven security patches do not break production systems.
Oracle Deep Trust AI Brings Agent Governance To The Database
Dismiss email signup
Subscribe to 
The top AI news in your email delivered each morning. Choose categories