Tag: Cybersecurity

Astra tests OpenAI’s own red line for cyber capability

OpenAI says internal tests can no longer rule out a Critical-level cyber rating for its Astra model.

Jamie Dimon recruits CEOs for cross-industry AI risk body

JPMorgan's chief is personally signing up more than 40 companies to a revamped Alliance for Critical Infrastructure focused…

Obsidian hits unicorn status with $85M for AI agent guardrails

Obsidian Security banks $85 million in a Series D that values the AI agent governance startup at $1.1…

Agent security startup Zenity banks $125M for the copilot era

Security firm Zenity raises $125M to govern enterprise AI agents that can access data and act on their…

Horizon3 lands $250M to keep attacking before attackers do

Autonomous pentesting firm Horizon3 banks $250M to expand its AI-driven NodeZero platform past a $2B valuation.

Okta snaps up Permiso in $200M bet on machine identity

Okta is buying AI identity security startup Permiso for nearly $200M.

Claude models slipped online and hacked three real firms

A misconfiguration let three Claude models reach the live internet during security tests, where they breached real organizations.

Microsoft’s new cyber model finds 96 percent of real bugs

Microsoft's first cybersecurity AI model catches 96 percent of real-world bugs at half the cost of frontier systems.

Industry coalition unites behind open AI security

NVIDIA and 50+ partners launch an open-source coalition to defend critical infrastructure from AI-powered threats.

Hacked Suno leaks 55M accounts and payment data from AI music platform

A data breach at AI music startup Suno exposed 55 million user accounts plus Stripe payment records, while…

AI models outgrow existing cybersecurity benchmark tests

Frontier AI hacking capabilities are saturating benchmarks within weeks, forcing a government and industry rethink.

AI Reasoning Models Vulnerable to Overthinking-Based Denial-of-Service Attacks

Researchers have discovered that AI reasoning models can be weaponized through overthinking, turning illogical prompts into denial-of-service attacks…