Abnormal AI’s 2025 updates expand human threat coverage

At a glance

What this is: NHI Mgmt Group's analysis, drawing on Abnormal AI's blog post: 2025 update extends human threat coverage beyond email into calendar invites, Microsoft Teams, and misdirected send prevention, while reporting a 30% reduction in missed phishing detections.

Why it matters: IAM and security teams need to treat collaboration platforms, user behaviour, and identity context as part of the attack surface, not as separate problems from email security.

Context

The security gap here is not only email filtering, but the mismatch between how attackers operate and how many human-threat controls are still organised. Human-targeted attacks now move across inboxes, calendars, and collaboration tools, so controls that only inspect mail content miss part of the path.

For identity and access teams, the important issue is that user behaviour and permission context now matter as much as message analysis. When a phishing lure lands in Microsoft Teams or a risky send action happens inside email, the control point has shifted from perimeter filtering to behavioural detection and remediation.

Key questions

Q: How should security teams defend against phishing when attacks move beyond email?

A: Security teams should shift from inbox-centric prevention to browser-aware detection, stronger app-consent governance, and post-login monitoring. Non-email phishing often bypasses sender controls entirely, so the browser becomes the key enforcement point. Teams also need identity telemetry that can identify session theft, malicious redirects, and risky OAuth grants after the lure succeeds.

Q: Why do calendar invites and chat messages bypass traditional phishing controls?

A: They bypass traditional controls because those controls were built around email-centric patterns and often assume lures arrive as messages with obvious malicious content. Trusted collaboration channels carry social context that lowers user suspicion, so attackers can exploit normal business workflows rather than trigger obvious inbox warnings.

Q: What are the signs that human-threat detection is not keeping up?

A: Common signs include missed detections on internal impersonation, poor coverage for Teams or calendar-based lures, and an overreliance on static indicators that generate either too many false positives or too many blind spots. If your controls work only when the attack looks like classic phishing email, coverage is already behind the threat.

Q: When should organisations add pre-send checks for misdirected email?

A: Organisations should add pre-send checks when sensitive information is routinely shared by email and the risk of wrong-recipient exposure is material. Pre-send intervention matters most where accidental disclosure would create compliance, confidentiality, or reputational harm, and where post-send containment would be too late to prevent the impact.

Technical breakdown

Why behavioural models reduce missed phishing detections

Behavioural detection differs from rule-based filtering because it looks for patterns in sender behaviour, message intent, language use, and relationship context rather than relying only on indicators already known to be bad. Abnormal AI says its models now use 50% more features, which suggests a broader signal set for distinguishing benign communication from impersonation and social engineering. The practical value is not merely more alerts, but better separation between malicious and legitimate activity across changing attack styles.

Practical implication: evaluate whether your human-threat stack can use behavioural context, not just static indicators and policy rules.

Why calendar invites and Teams expand the attack surface

Calendar invites and collaboration messages are attractive because they arrive through trusted internal workflows and often bypass the scrutiny applied to traditional phishing email. Once the lure moves into meeting tools or chat, the attack is no longer confined to the mailbox, so mailbox-only detections become incomplete by design. Abnormal AI's addition of calendar invite attack remediation and Microsoft Teams threat alerts reflects this channel shift, where the user experience itself becomes the delivery mechanism.

Practical implication: include collaboration platforms in phishing response workflows instead of treating them as out-of-scope communications channels.

How misdirected email prevention changes data loss control

Misdirected email prevention is a human-error control, not an attacker control. It uses behavioural context and identity awareness to identify when a user is about to send sensitive information to the wrong recipient, which matters because many exposure events begin as routine mistakes rather than malicious compromise. In identity governance terms, this is an example of controls moving closer to the action moment, where recipient context can prevent accidental disclosure before the message leaves the tenant.

Practical implication: decide whether your data-loss controls can intervene before send, not just after content leaves the mailbox.

NHI Mgmt Group analysis

Human-threat defence is becoming a multi-channel identity problem. Email remains important, but the control boundary now extends into collaboration systems where users act with trusted identities and attackers exploit that trust. Once calendar invites and Teams become delivery paths, the issue is no longer only message inspection. Practitioners should treat human identity risk as a cross-channel governance problem, not an email-security feature set.

Behavioural context is replacing static indicators as the decisive control plane. The article's 50% more features and 30% reduction in missed detections point to a broader security truth: modern phishing defence depends on richer context, not just more signatures. That aligns with human IAM realities where identity, communication pattern, and action intent matter together. The practical conclusion is that detection quality now hinges on context quality.

Collaboration tooling has become part of the user-authentication trust chain. When malicious content arrives in Teams or via calendar invites, the user's trust in the channel is part of the attack. That does not make the collaboration tool itself the vulnerability; it makes trust transfer the vulnerability. Security teams should re-evaluate where they assume trust because of channel familiarity.

Misdirected email prevention is an overlooked governance control for human data exposure. Many programmes focus on external attackers while ignoring the routine error path that moves sensitive data to the wrong person. This is a governance gap because the failure occurs before transmission, where downstream DLP alone may be too late. Identity and access leaders should view pre-send intervention as part of human data protection, not a niche productivity feature.

From our research library:

• Security researchers tracked consent phishing campaigns affecting 900 tenants and 3,000 user accounts in 2025.

What this signals

Channel expansion is the real story. Security programmes that still centre human-threat defence on email alone will miss the operational shift already underway. Teams should plan for response ownership, detection logic, and user coaching to extend into collaboration tools because that is where attacks increasingly land.

Behavioural context will keep displacing static indicators. As attackers personalize lures and move faster across channels, the quality of the context feeding detection becomes a governance issue, not just a tuning issue. Practitioners should expect more of the control burden to move from inbox filtering to identity-aware interpretation of user behaviour.

For practitioners

• Extend human-threat monitoring beyond email Include Teams, calendar invites, and other collaboration surfaces in phishing detection, triage, and remediation workflows so lures are not handled as mailbox-only events.
• Test behavioural detection quality Measure whether your current controls reduce missed detections without driving false positives higher, especially for internal impersonation, compromised senders, and brand abuse.
• Add pre-send controls for accidental disclosure Use recipient-aware checks that stop misdirected email before transmission when the message contains sensitive information or the recipient context looks inconsistent with past behaviour.
• Separate collaboration-channel response from mailbox response Define who can review, quarantine, and remediate malicious content in Teams and calendar systems so response ownership follows the channel where the exposure occurs.

Key takeaways

• Human-targeted attacks are no longer confined to email, and controls that stop at the inbox will miss lures delivered through collaboration tools and calendar workflows.
• Behavioural models are being used to reduce missed detections without increasing false positives, which points to context quality as a key control variable.
• Pre-send intervention for misdirected email matters because some of the highest-value data exposures begin as ordinary user mistakes rather than overt attacks.

Key terms

• Behavioral Detection: A monitoring approach that looks for unusual activity rather than relying only on static inventories. For SaaS integrations, it detects drift in token use, data movement, timing, and endpoint behavior so teams can spot compromise, misuse, or automation that no longer matches its expected pattern.
• Collaboration Channel Abuse: The misuse of trusted collaboration tools such as Teams, calendar systems, and internal messaging to deliver phishing or social engineering payloads. These channels amplify trust, reduce user suspicion, and can carry identity-related abuse into everyday workflows.
• Misdirected Email Prevention: Misdirected Email Prevention is a control that detects likely email misdelivery before a message leaves the tenant. It uses behavioral and contextual signals to identify recipient errors, then routes the message to quarantine or prompts the sender for review. The goal is to prevent accidental outbound data loss without blocking normal work.
• Identity context: The entitlement, ownership, and purpose information that explains why an action occurred and whether it was expected. For security operations, identity context turns raw alerts into decisions by showing which human or non-human identity acted and what it was allowed to do.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.