Security advice has spent a decade telling people not to click the link. That advice still matters, but it is no longer covering the ground it used to. A run of recent disclosures shows attackers routing around the human step entirely: no bad click, no suspicious attachment, no moment where a distracted account manager could have caught it. The compromise happens in the infrastructure underneath the work, in places most marketing and creative teams never look because nobody ever taught them to.
What "no click required" actually looks like
The clearest example is the WeChat zero-click worm that took over accounts on both iPhone and Android through an incoming call. No message to open, no file to run, just a call arriving. That is the shape of the threat now: the entry point is a normal, expected interaction with the platform, not an anomaly a trained user could learn to spot.
The same shift is happening on the tooling side of attacks. Hackers are now building dedicated AI frameworks for credential theft, which means the theft itself is being automated and scaled the way legitimate software gets built, with reusable components rather than one-off scripts. A FreeIPA flaw chain went further still, letting anonymous clients create reusable administrator credentials, no valid login attempt anywhere in the chain for a security team to flag.
None of this requires the target to make a mistake. It requires the target to have a system running.
The parts of the stack nobody audits
Most agencies can tell you who has access to the CMS login. Far fewer can tell you what has silent, machine-level access underneath it. That is the gap this run of disclosures keeps landing on.
- Adobe fixed a critical Magento zero-day that was being exploited to backdoor servers outright, which matters directly to any agency running client storefronts on Magento rather than a hosted platform.
- MikroTik patched critical flaws that were being chained together to hack routers, the kind of network hardware that sits invisibly behind office wifi and client VPN connections and is rarely on anyone's patch calendar.
- N-able patched a critical zero-day in N-central, its remote monitoring and management platform, the exact category of tool that managed service providers use to reach into dozens of client environments at once.
- A webinar on Google Workspace flagged what it called forgotten access, the kind that can lead to a breach on its own, which is a direct hit for any agency whose entire client collaboration runs through shared Drives and delegated admin roles nobody has revisited since setup.
Every one of these is a piece of infrastructure that a marketing team relies on daily without ever thinking of it as an attack surface. The Magento store, the router, the RMM tool, the Workspace admin panel: none of them require a click from anyone on the creative team to be exploited, and none of them show up when the audit question is "who has been trained on phishing."
What happens after the silent entry
The downstream damage from this kind of access is not abstract. Mathspace, an education technology company, had a data breach that exposed over 1 Million People. A Bavarian municipal utility had its systems encrypted by ransomware, the kind of operational hit that starts with exactly the sort of quiet infrastructure access described above. French prosecutors confirmed the arrest of a suspected hacker behind a tax cyberattack under the ZeroBytes name, evidence that this activity is being investigated and prosecuted, not treated as background noise.
And the money trail is real too. The party ended for crypto scammers who went on a spending spree after a $240 Million Bitcoin theft, which is worth sitting with: theft at that scale is now routine enough to generate its own aftermath story about the spending habits of the people who pulled it off.
New surfaces worth watching
Two other items in this run point at where this goes next. Researchers have been looking at stealing AI reasoning traces, meaning the internal working of an AI system, not just its output, is becoming something worth taking. Any agency running AI tools against client data or strategy should treat the reasoning layer, not just the final deliverable, as something with a security boundary around it.
Separately, a piece on reaching 1 Billion Build Manifests is a reminder of the sheer scale modern software supply chains now operate at. If your creative stack includes anything built through a CI/CD pipeline, whether that's a client site, a plugin, or an internal tool, that pipeline is part of your attack surface whether anyone signed off on it as such.
Microsoft's own Windows Server 2025 changes have been causing app crashes, which is a smaller story but a useful one: patching and hardening a system can break things that were working, and that friction is exactly why teams put off updates on infrastructure they don't think of as theirs to maintain.
What an audit needs to actually check
The practical response is not another phishing training session. It is a list of the systems in your stack that have standing access without a human login event attached to each use, and a check on when each one was last reviewed.
- List every RMM, monitoring, or remote-access tool a vendor or agency partner has installed into your environment, and confirm who still needs it.
- Check Google Workspace (or whichever suite you run) for delegated admin roles, forgotten service accounts, and old app authorizations nobody has revisited.
- Confirm patch status on anything self-hosted: Magento stores, routers, on-prem identity tools. If it's not on a managed platform, someone has to be the one watching for the next zero-day.
- Ask what build pipeline your dev team or agency partner uses for client sites and tools, and whether that pipeline itself has ever been audited.
- If you're using AI tools against client material, ask the vendor what happens to the reasoning traces, not just the outputs, and where they're stored.
None of this is glamorous work, and none of it produces a case study. But the incidents above share one property: by the time anyone noticed a click, it was already too late to matter.