• Say a few words, and a rigged AI asks for $5,000
• AI coding boom exposes cracks in cloud development
• AI agent observability is struggling to keep up with hallucinations
• Chrome puts insecure sites behind a permission prompt
• New PhaaS Platform Uses AI to Craft Phishing Lures and Mask Malicious Activity
• eMudhra security blunder left trusted CA signing keys exposed to anyone
• Symantec Uncovers AI Agent Hijacking Models for Autonomous Cyber-Attacks
• NVIDIA unveils security architecture that can stop AI agents crossing set boundaries
• PHP2Shell: Critical PHP vulnerabilities patched, update now!
• Ransomware Attacks Rise as Rival Gangs Vie For Market Supremacy
• Ransomware gangs have been exploiting Palo Alto GlobalProtect VPN flaw since June
• SafePay Has Compromised 600 Critical Infrastructure Victims Worldwide, CISA Warns
• Critical PHP Vulnerabilities Leaked Ahead of Scheduled Security Patches
• Children’s Data Stolen After Ransomware Attack Compromises ‘Password1’ Account
• Head of AI Security Roles Surge in Large Organizations
• Researchers Warn OpenCode Quietly Patched Critical RCE Flaw, Leaving Users At Risk
• AI Agent Adoption Creates Risk as Oversight Struggles to Keep Pace, Splunk Finds
• Siemens Customers Targeted by ClickFix Attackers Exploiting DLL Side-Loading Flaw
• Exploited Vulnerabilities Typically Take Three Months to Patch, Moody’s Finds
• Conduent Agrees to Pay $80m to Settle Data Breach Class Action
• OpenCode Breach Could Affect OpenAI, Anthropic and Google
• ConnectWise Shuts Off Cloud Services After ‘Unauthorized Access’ to Microsoft 365 Tenants
• US Cybersecurity Job Postings Fall 26%, Hitting Four Year Low

Researchers at Imperial College London tampered with speech-to-text models so they transcribed most speakers accurately and misbehaved for one chosen person. When that person spoke, the models produced output the attacker had picked in advance. In one test, that output was a request to send $5,000, and AI agents reading the transcript made the transfer in a sandboxed banking environment. The attacker needs access to the model only while it is being fine-tuned, through a …