Security.io Intelligence DeskFriday, 4 September 2026
Independent analysis
for security executives
The Security.io DailyThe Weekday Intelligence Edition
Free to readers
Supported by underwriters
Intelligence desk

AI Security

A permanent file of Security.io’s selected briefings tagged to this executive theme. It reflects the publication’s curated coverage, not a census of all global activity.

2026-09-04ASCII smuggling moves from prompt injection into phishing evasionEmail Security
2026-09-02FSB reframes frontier AI as systemic cyber-resilience riskAI Security
2026-09-02Langflow exploitation shifts AI tooling into credential containmentAI Security
2026-08-31ServiceNow’s three unauthenticated critical flaws put deployment ownership under scrutinyCloud Security
2026-08-21Active Siemens S7 targeting turns PLC exposure into a safety decisionOperational Technology
2026-08-21Actively exploited MLflow flaw can expose cloud credentialsVulnerability Management
2026-08-20CISA adds MLflow SSRF flaw to exploited-vulnerability catalogueVulnerability Management
2026-08-20Federal agencies warn of active AI-assisted targeting of Siemens S7 PLCsOperational Technology
2026-08-18Active Ray exploitation turns developer AI environments into an incident questionAI Security
2026-08-17Claude outage shows AI continuity must be service-specificResilience
2026-08-13NIST asks how the NVD should operate in the age of AIVulnerability Management
2026-08-12AI vulnerability artefacts need semantic verification, not a successful runAI Security
2026-08-11GPT-5.6-Cyber turns provider access into an enterprise control-plane decisionAI Security
2026-08-10Self-evolving agent skills create a trajectory-poisoning control gapAI Security
2026-08-07OpenAI’s Black Hat timeline moves the first containment failure to 26 May 2026AI Security
2026-08-06Poisoned replay data can silently break adaptive intrusion detectionAI Security
2026-08-06PURPOSE shows how RAG poisoning can evade contradiction checksAI Security
2026-08-05Agent frameworks need containment after prompt injection succeedsAI Security
2026-08-04Reported AI-managed proxyjacking campaign needs verification, not dismissalAI Security
2026-08-03EU AI Act transparency enforcement begins, shifting AI inventory from programme work to evidence obligationRegulatory
2026-08-03OpenAI–Hugging Face incident makes AI evaluation containment a privileged-system decisionAI Security
2026-07-31Claude evaluations reached real production systemsAI Security
2026-07-30OpenAI evaluation incident expanded to four external service accountsAI and Cloud Security
2026-07-29OpenAI update identifies Artifactory escape path in Hugging Face intrusionAI Security and Incident Response
2026-07-27Recovered intrusion logs show an AI agent executing unattended post-exploitation tasksAI and Emerging Threats
2026-07-23AI cyber evaluation crossed containment and reached Hugging Face productionAI and Emerging Technology
2026-07-15AI-assisted discovery is exposing a capacity mismatchSecurity engineering
2026-07-15An AI-backed vulnerability clearinghouse will not solve enterprise prioritisationPublic policy
2026-07-14Dialogflow flaws show that conversational agents have control planesAI security