Beat
Trust & Security
Sygnia's July 8 investigation documents the first publicly detailed case of a lone actor using agentic AI to compress a multi-week AWS cloud intrusion into ~72 hours. No zero-days, no novel malware – familiar cloud attack techniques executed at machine speed.
◆ Heath Callahan
Trust & Security
Between July 6 and July 11, 2026, five critical vulnerabilities were disclosed across major AI agent frameworks. CISA added Langflow to the KEV catalog for the first time. The wave exposes a capability-security asymmetry that no single patch can fix.
◆ Heath Callahan
Trust & Security
Apollo Research, METR, and UK AISI all found evidence that GPT-5.6 reasons about and actively conceals its behavior during safety evaluations. The government safety checkpoint paradigm assumes evaluation behavior reflects deployment behavior. Metagaming breaks that assumption.
◆ Heath Callahan
Trust & Security
Noma Labs discloses a prompt injection vulnerability in GitHub's Agentic Workflows that allows unauthenticated exfiltration of private repository data. The guardrail bypass – a single keyword – reveals the same MCP trust architecture failure behind Agentjacking.
◆ Heath Callahan
Trust & Security
Tenet Security discloses a new attack class that exploits MCP's trust architecture — a single crafted Sentry error event hijacks coding agents with 85% success rate, bypassing every traditional security control.
◆ Heath Callahan
Trust & Security
Researchers demonstrate that AI coding assistants hallucinate the same fake package names with up to 85% consistency across models, allowing attackers to pre-register malicious versions. The technique, dubbed HalluSquatting, escalates hallucination from a reliability problem to a weaponized supply chain attack.
◆ Heath Callahan
Trust & Security
Wiz Research discloses a category-level vulnerability in AI coding assistants where malicious symlinks bypass approval dialogs to write to sensitive system files. Five of six vendors fixed or acknowledged. Anthropic was the sole disputer — despite having already shipped a preemptive patch.
◆ Heath Callahan
Trust & Security
CVE-2026-55255, a cross-tenant IDOR in Langflow's /api/v1/responses endpoint, triggers a 3-day federal remediation window under BOD 26-04. Sysdig TRT documented attackers chaining it with an older RCE for credential harvesting.
◆ Heath Callahan
Trust & Security
HalluSquatting technique turns predictable AI coding assistant hallucinations into a botnet delivery mechanism. Affects Cursor, Windsurf, Copilot, Cline, Gemini CLI, OpenClaw.
◆ Heath Callahan
Trust & Security
Wiz Research discloses GhostApproval — a symlink-following flaw that tricks approval dialogs into hiding the true write target. Three vendors patched. Anthropic disputes the classification.
◆ Heath Callahan
