Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Salt Labs recently showed that a single email could hijack the AI agent platform Manus and reach a victim’s connected accounts. The full technical breakdown, with complete evidence, payloads, and screenshots, is in our research team’s write-up, and Dark Reading first reported the finding in an exclusive. This post is the shorter version: what the finding means and what it tells every organization now deploying AI agents. For the full technical detail, read the Salt Labs research blog.

New training on using AI safely and spotting AI-powered threats arrives this Cybersecurity Awareness Month Attackers are moving faster than ever, weaponizing newly discovered vulnerabilities before defenders can patch them and building new types of malware that signature-based defenses can’t catch.

Attacking a CI/CD pipeline used to require a specialist who understood Git internals, cloud identity, and the way build runners handle secrets. That is no longer true. The barrier to entry for an advanced attack has dropped to the level of writing prompts in English. Automation itself is not new to DevOps, but AI has raised its ceiling on both sides of the fence.

An agentic SOC uses AI-powered investigation, analysis, and automation to gather data, connect activity, and handle repetitive investigative work. Analysts remain in control and focus on judgment, prioritization, and response. This addresses the need for machine-speed security operations as AI agents gain autonomy and Tier 1 access to systems. Spend five minutes watching a security analyst at work and the “AI will replace analysts” headline starts to sound out of touch.

CVE-2026-86858 is an improper access control flaw in the ServiceNow AI Platform, classified as CWE-284. ServiceNow describes it as an unauthenticated privilege escalation reachable through GraphQL. In certain circumstances an unauthenticated user can create, modify, or delete instance data beyond what was intended. The vulnerability carries a CVSS v4.0 base score of 8.7 (High). Attack vector is network, attack complexity is low, and neither privileges nor user interaction are required.

A model produces a figure, an event happens, and somebody asks whether the figure was right. It is the obvious question and it has almost no published answer, because the comparison is harder than it looks. ‍ A single realized loss cannot falsify a distribution. If a model puts a one percent chance on exceeding a threshold and the threshold is exceeded, the one percent case occurred, which is what the model said would sometimes happen. ‍

You’ve seen the demos. You’ve run the pilots. Maybe you even deployed something that works, but is it scaling? Are your users utilizing it on a daily basis, or as intended? Or is it quietly gathering dust next to the last three tools that were supposed to change everything? Scaling AI the way most vendors describe it—connecting more tools and piping more data into more systems—doesn’t lead to progress or productivity. Rather, it results in additional risk.

The speed of AI integration is rapidly outpacing corporate governance capabilities. Generative AI is now embedded in development workflows and business applications, while agentic AI can make decisions and take actions with limited human intervention. That governance gap is becoming harder to ignore. Gartner predicts that more than 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value, or inadequate risk controls.

AI coding agents produce authorization logic that compiles, passes review, and enforces the wrong policy. Broken access control ranks first in the OWASP Top 10:2025, where 100% of applications tested showed some form of it, across 1,839,701 recorded occurrences, the highest count of any category on the list. One part of that category is also the part that pattern-based scanning was never built to reach.

AWS Dogwood brings stateful authorization to AI agents. Learn how it fits with workload identity, AuthZEN, IAM, and the move away from long-lived credentials.