TL;DR: Browser-extension phishing, real-time credential harvesting, and AI credential abuse are recurring attack patterns in Anomali’s Cyber Watch roundup, with attackers exploiting trusted login flows, compromised accounts, and exposed secrets to reach downstream systems. The lesson is that identity controls must cover how credentials are captured, used, and abused, not just how they are issued.
At a glance
What this is: Anomali’s Cyber Watch roundup shows browser-extension phishing, live credential harvesting, and AI-related credential abuse are converging into a practical identity risk pattern.
Why it matters: It matters because identity teams need to govern the moment credentials are captured and reused, not just the policies that issue access in the first place.
Context
Browser-extension phishing, live login interception, and AI-assisted credential abuse all exploit a common weakness: users and systems still trust the visual signs of a legitimate sign-in flow. In practice, that means the browser, the authentication prompt, or the delegated workflow becomes part of the attack surface rather than a trusted boundary.
For IAM and NHI programmes, the important issue is not whether the abuse is human, automated, or AI-assisted. The security gap is the same when an attacker can intercept credentials, tokens, or approvals at the point of use and turn a normal login sequence into a takeover path.
Key questions
Q: What breaks when browser extensions are allowed to modify login pages?
A: Browser extensions can replace or overlay a legitimate login surface while the address bar still shows the correct site, which defeats visual trust cues. That breaks assumptions built into user training and many phishing defenses. The control failure is usually too much extension privilege and too little enterprise oversight of what extensions can read, inject, or intercept.
Q: Why do live phishing and one-time passcodes still lead to account takeover?
A: Because the attacker captures credentials and the passcode while the session is active, then replays them before the code expires or the user notices the deception. This defeats controls that depend on delayed verification or user-in-the-loop prompts. Phishing-resistant methods reduce that replay window by binding authentication to the real session.
Q: Where do delegated AI workflows create NHI governance failures?
A: They fail when the workflow, plugin, or agent can use credentials beyond the exact task it was meant to complete. At that point the organisation has effectively created a non-human identity with unclear authority and poor lifecycle control. The risk is not just theft, but credential use in unintended contexts that security teams do not routinely review.
Q: Should teams prioritise browser controls or authentication controls first?
A: They should treat both as linked, but start with the control that closes the active abuse path in their environment. If phishing is succeeding through browser manipulation, extension governance matters immediately. If login replay is the primary issue, phishing-resistant authentication should move first. The right order depends on where the attack is winning today.
Technical breakdown
How browser extensions turn trusted pages into phishing surfaces
Browser extensions can modify page content after the page has loaded, which means the user may still see the real domain while interacting with attacker-controlled content. A malicious extension can inject overlays, full-screen iframes, or lookalike forms that capture credentials without breaking the browser’s visible trust cues. The result is not a simple spoofed site, but a compromised interaction layer sitting on top of a real one. This matters because traditional phishing filters and URL checks can miss attacks that happen inside the browser runtime rather than at the network edge.
Practical implication: Treat extension permissions and post-load page modification as a governed control surface, not a user convenience.
Why live credential harvesting defeats delayed verification models
Real-time phishing changes the economics of credential theft. Instead of stealing a password for later use, the attacker captures credentials and one-time passcodes during the authentication session and immediately replays them before the challenge expires. That collapses the value of controls that assume human review or delayed response. Once the attacker is in the live login flow, the main question becomes whether the identity system can distinguish legitimate use from a proxied session in progress. Phishing-resistant authentication narrows that window by removing the attacker’s ability to reuse what was just captured.
Practical implication: Prioritise phishing-resistant authentication where approval-based login flows still allow immediate replay.
How delegated and AI-assisted access creates credential abuse risk
AI-related credential abuse often emerges when a delegated workflow, plugin, or agent is given access that outlives the task it was meant to perform. If secrets, tokens, or approval paths are available to the wrong runtime at the wrong moment, the identity boundary shifts from the human to the system acting on their behalf. That creates a non-human identity governance problem even when the initial interaction looks human-led. The operational risk is not only theft, but ambiguous authority: a valid credential being used in an unintended execution context.
Practical implication: Scope delegated access to the shortest possible task window and inventory every AI or automation path that can use credentials.
Threat narrative
Attacker objective: The attacker aims to take over trusted accounts quickly enough to bypass normal authentication checks and reach downstream systems before the stolen credentials lose value.
- Entry occurs through browser-extension abuse or live phishing that preserves the appearance of a legitimate login flow.
- Credential harvesting happens in real time as passwords, one-time passcodes, or approvals are captured during the session.
- Escalation follows when the attacker reuses the captured identity material immediately to access downstream systems and accounts.
- Impact is unauthorized access, extortion pressure, or broader identity compromise across connected services.
Breaches seen in the wild
- CoPhish OAuth phishing via Copilot Studio: Datadog showed Copilot Studio agents on a Microsoft domain can front OAuth consent phishing and forward stolen tokens; no victims reported.
- Canvas Instructure Data Breach: ShinyHunters exploits Canvas LMS platform to expose millions of student records via third-party NHI credential abuse.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Browser-layer trust has become part of identity security: once an extension can alter what a user sees after page load, the browser itself becomes an identity control boundary. That is a governance problem, not just a phishing problem, because the user may authenticate against an interface the organisation never intended to trust. Practitioners should treat browser integrity and extension scope as part of access control design.
Real-time credential capture breaks the assumption that authentication is a one-time event: the moment attackers can harvest and replay credentials inside the same login session, delayed detection and manual review lose much of their value. The control gap is not weak passwords alone, but the trust model that assumes a credential can be safely reviewed after use. Teams need to move protection to the point of issuance and use, not after compromise.
AI credential abuse is an NHI governance problem when delegated runtime access outlives the task: if automation, extensions, or agents can use credentials outside their intended context, the organisation has created a non-human identity with ambiguous authority. That is the same governance failure whether the actor is a bot, an extension, or an AI workflow. The practical lesson is that identity lifecycle rules must follow runtime use, not just account creation.
Credential-use visibility is the named control gap: these campaigns exploit the gap between credential issuance and credential use, where controls often become less effective once the login flow begins. Organisations that only monitor enrolment, password policy, or account issuance miss the actual abuse point. The right question is whether identity governance can see and constrain how credentials behave during live sessions and delegated workflows.
What this signals
Browser trust is now an access-control problem: organisations that still separate browser management from identity governance are missing how phishing actually reaches users. Extension permissions, page injection, and visual trust cues now sit inside the authentication path, so access policy has to cover the browser runtime as well as the login system.
A second implication is that AI and automation teams need the same discipline they already expect from IAM teams: tight scoping, revocation, and explicit ownership. When delegated access can be reused outside the intended task, credential abuse stops being an edge case and becomes a governance failure.
The practical direction is toward issuance-time control, runtime monitoring, and strong authentication methods that cannot be trivially proxied. Programmes that rely on user recognition of suspicious pages, delayed review, or broad delegated access will keep absorbing the same class of abuse under different packaging.
For practitioners
- Enforce browser extension allow-listing Restrict extensions to an approved set, review requested permissions, and remove anything that can alter page content or capture form input.
- Prioritise phishing-resistant authentication for critical access Move high-risk users and privileged workflows to methods that cannot be replayed from a captured prompt or one-time code.
- Review delegated AI and automation access paths Inventory where scripts, agents, or integrations can use credentials, then shorten task scope and revoke any standing access that is not operationally required.
- Instrument suspicious login journeys as security events Correlate unusual approvals, support-themed contact, and rapid token use so that authentication abuse is visible as an access issue, not just a fraud signal.
Key takeaways
- Browser-extension phishing can preserve the appearance of a legitimate login flow while the content on the page is attacker-controlled.
- Live credential harvesting is dangerous because attackers can replay credentials and one-time passcodes before the authentication window closes.
- Identity programmes need controls that govern browser integrity, authentication method strength, and delegated runtime access at the point of use.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10 and MITRE ATT&CK address the attack and risk surface, while NIST CSF 2.0 and CIS Controls v8 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-04 — Insecure Authentication | Live phishing and replay abuse target authentication flow trust, not just secret storage. |
| NHI-10 — Human Use of NHI | Delegated automation and AI use of credentials creates human-to-machine reuse risk. | |
| Recommendation — Use NHI-04 to harden authentication paths against proxying, replay, and user-mediated capture. Apply NHI-10 to remove human credentials from workflows used by scripts, bots, or agents. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The article centres on credential capture followed by immediate account use. |
| Recommendation — Map these campaigns to TA0006 and TA0008 to prioritise detection around credential capture and reuse. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The issue is governed access at the point of use, especially for browser and delegated workflows. |
| Recommendation — Apply PR.AA-05 to limit which sessions, apps, and workflows can use sensitive credentials. | ||
| CIS Controls v8 | CIS-5 — Account Management | Compromised accounts and unmanaged browser tooling both point to account lifecycle weakness. |
| Recommendation — Use CIS-5 to inventory accounts and remove unmanaged access paths that enable credential abuse. | ||
Key terms
- Browser-based phishing: Browser-based phishing is phishing that executes through the web browser rather than the inbox, often using redirects, malicious sites, consent prompts, or extensions. It matters because the browser is where identity, application access, and session state intersect.
- Phishing-Resistant Authentication: Phishing-resistant authentication proves identity without relying on a user to approve a prompt or reveal a reusable secret. It typically binds access to a device, key, or cryptographic proof that an attacker cannot easily reuse or coerce. This approach reduces reliance on human judgment at login time.
- Delegated credential abuse: Delegated credential abuse occurs when a workflow, automation path, or agent uses credentials beyond the task or context it was intended to support. The security problem is not only theft, but misuse of valid access in the wrong runtime or business process.
- Non-Human Identity Governance: Non-human identity governance is the practice of managing, controlling, and auditing every machine identity across its full lifecycle. It covers service accounts, API keys, tokens, certificates, and AI agent credentials, ensuring each has a defined owner, scoped privilege, rotation schedule, and revocation path. Without governance, NHIs accumulate silently and become the primary attack surface in cloud and automated environments.
Deepen your knowledge
NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course. It is designed for practitioners who need to connect access governance, credential control, and runtime trust across human and non-human identities.
Published by the NHIMG editorial team on August 2, 2026.
Updated on October 12, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org