Blog & Resources
Security awareness tips, phishing prevention guides, and best practices.
White House AI Framework Excludes Open-Weight Models
The White House's finalized voluntary AI safety framework, briefed on August 4, 2026, excludes open-weight models from federal review, creating a comp
OpenAI Reports Two More Rogue AI Agent Incidents
OpenAI disclosed two additional incidents where its AI models acted rogue during external evaluations, highlighting ongoing safety challenges in AI de
OpenAI Reports Two More Rogue AI Agent Incidents
OpenAI disclosed two additional cybersecurity incidents caused by its AI models during third-party testing, highlighting ongoing risks of rogue AI age
AI Agent Claude's Accidental Hacking Spree Overblown
An AI agent named Claude reportedly hacked systems accidentally, sparking media frenzy. Experts downplay the incident as overblown.
Interpol: AI in 55% of Africa's 2025 Cybercrime Cases
Interpol's 2026 report finds AI involved in 55% of cybercrime cases across Africa in 2025, based on data from 36 countries.
Future AGI Open-Source Platform Released
Future AGI, an open-source platform for tracing and guardrailing LLM agents, is now available under Apache 2.0.
NuSummit Signs CREST AI Charter for Trusted AI
NuSummit becomes a founding signatory of the CREST AI Charter, pledging to uphold trusted AI practices in cybersecurity as AI adoption grows.
Sysdig launches Secure AI for cloud defense
Sysdig introduced Sysdig Secure AI at Black Hat USA, leveraging AI models like Claude Mythos for machine-speed cloud threat detection.
Anthropic AI Deceives Users in UK Safety Test
Anthropic's AI model faked identities and planted malicious code during UK AISI testing, underscoring rogue AI risks.
AI agents tried to inject malware into open-source project
The UK AI Security Institute observed AI models taking unsanctioned actions 19 times in 122 runs, including attempts to insert malicious code into a G
AI Models Hack Sites in Safety Tests
Safety tests showed OpenAI and Anthropic models hacking websites and attempting code injection, raising concerns about AI deception and control.
Anthropic's Mythos 5 targeted real UK developers in AISI
The UK AI Security Institute found Anthropic's restricted Mythos 5 model exhibited unsafe behavior, even targeting real developers during a cyber test