Microsoft has disrupted infrastructure associated with EvilTokens after obtaining authority from a US federal court. Axios reports that the service compromised more than 12,000 email inboxes across 10,000 organisations, including companies in finance, healthcare, property and education.

The reported technique used device-code phishing. This can trick a victim into authorising an attacker-controlled session, allowing the criminal to obtain tokens without stealing a conventional password. Once inside an email account, an attacker can impersonate executives, redirect payments, steal data or move deeper into connected systems.

The AI element matters because criminal services can automate targeting, message creation and operational support. EvilTokens reportedly charged an initiation fee and monthly subscription, lowering the expertise needed to run campaigns at scale.

A court-backed takedown removes infrastructure; it does not automatically invalidate every stolen token or eliminate copycat services. Organisations still need rapid token revocation, controls around device-code authentication, behavioural detection and strong monitoring of new application consent. The useful enterprise lesson is that identity has become the control plane for both human and AI-assisted attacks.

Agentic Exploits- Deterministic gates for a probabilistic problem
David Girvin, CEO and co-founder of Assury, joins Stewart Tinson to dig into what’s actually happening when agentic AI goes wrong, and why he thinks most of the industry is solving the wrong layer of the problem. David explains the difference between prompt-level exploits and execution-level ones, arguing that the real danger starts the moment an agent moves from generating text to calling tools: deleting databases, reading files, sending emails. He walks through real-world incidents, including a Mexican government breach chain that escalated from just over a thousand prompts to over five thousand AI-executed actions across multiple agencies before detection, and the UK AI Security Institute’s recent cyber evaluation, in which agents took unsanctioned action including fabricating identities to socially engineer a real GitHub maintainer. The conversation covers why David is sceptical of “guardrails” language and AI-governing-AI approaches, arguing that only deterministic, architectural controls can reliably constrain agent behaviour, alongside human review reserved for genuinely high-stakes actions rather than blanket approval fatigue. He breaks down credential starvation, session risk accumulation, and why classifier-based tools keep failing inconsistently on identical actions, pointing to a named frontier lab’s own zero trust paper as an example of the industry misjudging what actually works. Elsewhere, David discusses the exposed MCP server problem, the widening trust gap between small specialist security vendors and platform incumbents, and why he believes regulation, not product quality alone, is what finally drives enterprise security spend. He closes with the exploit that concerns him most for the year ahead: session-level, goal-directed deception with no attacker involved at all.
BrightTALK

Share this post
The link has been copied!