Attackers abuse enterprise AI tools for theft, C2, and agent hijacking
A fake Postmark MCP package siphoned mail after 15 clean versions, and SesameOp hid C2 in the OpenAI Assistants API.
A fake Postmark MCP package siphoned mail after 15 clean versions, and SesameOp hid C2 in the OpenAI Assistants API.
Opinion says AI agents reward-hack and follow hidden instructions unless hard guardrails limit their reach.
Gurucul released AI Risk and Response to link AI activity with identity and security telemetry.
ESET warns SMB AI agents face malicious skills and prompt injection while classic attacks still dominate breaches.
Microsoft patched CVE-2026-85889 (CVSS 10.0), a missing-auth flaw in Azure AI Foundry enabling unauthenticated network privilege escalation; no exploitation observed.
Microsoft patched a CVSS 10.0 missing-authentication privilege escalation flaw, CVE-2026-85889, in Azure AI Foundry, already fully mitigated server-side.
Microsoft patched 18 elevation-of-privilege and information-disclosure vulnerabilities across Azure and Copilot products; all fixed server-side with no exploitation observed.
Microsoft is still investigating a bug that makes Copilot buttons disappear in classic Outlook for affected M365 Copilot users.
Microsoft is redirecting Microsoft 365 and Teams web users to new *.cloud.microsoft domains, requiring firewall and proxy rule updates by early October.
Microsoft is investigating a Microsoft 365 Copilot access disruption tracked as incident CP1470554, with no root cause or recovery timeline published.
ConfusedPilot attack exploits Microsoft 365 Copilot's retrieval pipeline to expose confidential enterprise data in AI responses.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 1.7K readers · 3.5K views · 2.6K reads · all time 161K views