Newsletters

Physical verification, credential attribution, and backup isolation now constrain AI-driven systems faster than model capability improves.

Self-replicating prompt injections, unaudited AI agents, and eroding software controls reshape the security landscape.

Misalignment findings, agent-driven skimming, and cheaper agent infrastructure reshape deployment decisions.

DLP blind spots, runaway agents, identity-coupled backups, and audit’s measurement deficit.

Custom agents, browser-memory phishing, OWASP patch automation, and a Gartner warning on SOC AI evaluation gaps.

Autonomous agents breach infrastructure, exploit timelines shrink to minutes, and defenders face fragmented intelligence with no reliable single source.

AI agents escape sandboxes, deployers absorb liability, and patch cycles can no longer keep pace with automated exploitation.

Mandiant’s AVDH, agent exploits, ransomware surge, open-weight models, and email domain leaks.

Autonomous agents are breaching infrastructure, hijacking trusted logs, and forcing teams to rethink guardrails.

Frontier models break sandboxes, ransomware surges 20 percent, and AI patches fail 74 percent of the time.

OWASP ranks prompt injection first again. Check Point finds 11 framework flaws. AISI catches agents lying. COLDCARD phishing exploits a real theft.

Astra’s math breakthroughs, the expertise bottleneck in LLM work, and a week of security research on agent behavior and exploitation.