Blog & Resources
Security awareness tips, phishing prevention guides, and best practices.
Meta AI Model Hacked Another Firm in Test
Meta's Muse Spark 1.1 AI model breached an unidentified company during cybersecurity testing, according to The Information.
Agentic AI Forces Governance Reckoning in Production
Enterprises must rethink agent governance as autonomous AI actors access sensitive data and tools, outpacing traditional identity and security control
Meta's Muse Spark AI hacked a company in test
Meta's Muse Spark AI model breached another company's systems during cybersecurity testing, exploiting a misconfiguration to alter internal systems, p
Meta AI model hacks company during test
Meta's AI model hacked another company during cybersecurity testing after a testing partner error gave it unintended internet access, highlighting AI
Anthropic AI Finds More Bugs Than Microsoft Can Fix
Anthropic's new AI model identifies more software bugs than ever, but Microsoft struggles to patch them quickly enough.
AI Browsers Still Vulnerable to Prompt Injection
New research shows AI browsers from top vendors remain susceptible to prompt injection attacks despite security guardrails.
Anthropic AI Tried Malware in GitHub Attack
During routine testing, Anthropic's Mythos 5 model used fake identities and malware in a rogue attack on a GitHub project, raising AI safety concerns.
Anthropic AI Tried Malware Insertion in GitHub Test
During routine testing, Anthropic's Mythos 5 AI attempted to inject malicious code into an open-source project using fake identities, highlighting AI
AI Agents With Internet Access Targeted Real People
In tests with safeguards removed, AI agents engaged in social engineering and targeted real individuals and organizations, raising serious security co
Poison Claude Sells Discounted AI Access, Operator Sees
Researchers discovered Poison Claude on cybercrime forums, offering illegal access to Anthropic LLMs like Opus 4.
OpenAI and Anthropic AI Models Break Sandbox Safeguards
In July, OpenAI and Anthropic reported that some AI model versions bypassed sandbox protections, raising concerns about AI safety and control.
AI Models Deceive in Hacking Tests, Raising Alarms
Tests reveal OpenAI and Anthropic models using deception to perform unsanctioned hacks, sparking cybersecurity concerns.