Skip to content Skip to footer

180,000 Screens Now Get Audited by an AI: SignageOS Joins Anthropic’s Cyber Programme

The weakest point in a digital signage network is rarely the screen. It is the software that decides what the screen shows.

The announcement

SignageOS has joined Anthropic’s Cyber Verification Program, gaining access to the AI developer’s Claude Mythos model to strengthen security across its platform.

The company is using Anthropic’s technology to identify and address vulnerabilities in its cloud services, APIs and software supporting more than 180,000 connected devices across over 50 hardware platforms.

The Cyber Verification Program provides verified organisations with access to advanced AI models for defensive cybersecurity applications. Claude Mythos can analyse large codebases and identify subtle vulnerabilities with less manual guidance than earlier models.

SignageOS says it is incorporating the technology into its security processes, including checks of software releases before deployment, and plans to share relevant findings privately with hardware manufacturers when vulnerabilities involve their platforms, allowing partners to address issues before public disclosure.

Why device management is the real target

Digital signage networks increasingly operate as part of enterprise IT infrastructure, which makes the platforms controlling them potential security targets.

A compromised management system can affect screens across multiple locations, regardless of the hardware manufacturer – one credential, thousands of faces. The consequence is not a data breach in the conventional sense but something more public: a screen in a station, a mall or an airport showing whatever an attacker wants.

That has already happened. In October 2025, several Los Angeles Metro bus stop displays showed a false bomb warning following a breach of a third-party content system – an incident that demonstrated how quickly a content-layer compromise becomes a public safety problem.

Most signage deployments are architecturally fragile in the same way. Content management systems sit above device management platforms, which sit above hardware running a variety of operating systems, and each layer is often supplied by a different vendor with its own patching cadence. A vulnerability in any layer can be exploited through the one above it.

What an AI code auditor adds, and what it does not

Claude Mythos is being applied to code review at scale: analysing large codebases to find subtle weaknesses with less manual guidance than earlier models.

That is a meaningful capability shift for an industry that mostly cannot afford comprehensive security review. A signage software vendor with fifty engineers is not going to run continuous deep static analysis across every API and service by hand; automated review changes what is economically possible.

The limits are worth stating. An AI auditor finds patterns, not business logic flaws that depend on understanding what a system is supposed to do. It cannot verify that a vendor’s threat model is right, that a privileged action is properly authorised at the business level, or that a supply chain dependency has not been trojanised. It reduces the cost of finding known classes of weakness; it does not replace penetration testing.

SignageOS’s existing posture – SOC 2 Type II and ISO 27001 certification, annual independent penetration tests and mandatory multi-factor authentication – is the layer that matters more than any single tool. Anthropic’s model adds automated vulnerability detection on top of a programme that already assumes external verification.

The quotes and what they reveal

“The platform that manages a screen network has to be at least as secure as the network itself,” said Michael Žabka, SignageOS co-founder and CTO, adding that Claude Mythos helps identify weaknesses conventional security tools and internal teams might overlook.

“AI is reaching digital signage from two sides: onto the screens themselves, and into the way screen networks are run, to the point where teams can manage them by simply asking. None of that works unless the platform underneath is trusted,” said chief executive Stan Richter.

The second quote describes an operational reality most networks have not prepared for. If a signage platform accepts natural language commands – “put this campaign on all screens in the north zone after 6pm” – then the platform is executing privileged actions from ambiguous input. The attack surface moves from the screen to the prompt.

The disclosure question

The plan to tell hardware manufacturers privately about vulnerabilities in their platforms is the most consequential commitment in the announcement, and it is rarer than it sounds.

Multi-vendor signage deployments create a blame gap: when a flaw sits in a hardware platform’s firmware, the software vendor that found it has no commercial incentive to disclose, because disclosure implies its own product was exposed. Private pre-disclosure to the partner, before public reporting, is the behaviour that turns a vendor relationship into a security relationship.

Whether it happens consistently is the test. A single private notification is goodwill; a published disclosure policy with timelines is governance. The industry has neither at present, and buyers rarely ask.

What buyers should now demand

Procurement of signage platforms has historically been driven by compatibility, price per device and content feature lists. Security appears as a checkbox: does the vendor have ISO 27001.

The questions that would actually change outcomes are harder. Can the vendor show a software bill of materials for the device agent? How quickly are critical patches pushed to estate devices, and can a network operator force an update without physical access? What happens to credentials when a screen is decommissioned? Who is notified, and how fast, when a vulnerability is found in a third-party hardware platform?

Those questions are answerable, and the vendors that answer them well will find that security becomes a sales asset rather than a cost centre – particularly in transport, healthcare and finance, where a compromised public screen is a reputational event rather than an inconvenience.

The sector signal

The larger pattern is that signage is being absorbed into enterprise IT and judged by enterprise IT standards. Once a screen network authenticates against a corporate identity provider, reports into a security operations centre and runs on managed devices, it inherits the obligations of any other IT estate.

That is good news for the platforms that invested early and bad news for the long tail of small vendors whose products were designed before anyone thought a billboard needed a patching policy. Consolidation in signage software is likely to be driven less by features than by assurance.

Why it matters, in one line

Auditing 180,000 screens with an AI code reviewer matters less for the vulnerabilities it finds than for the precedent it sets – that a signage platform’s security programme is now a product feature, and the next breach will be judged on whether the vendor could have known.

Source: invidis / sixteen-nine, “Cybersecurity: SignageOS Uses Anthropic AI to Secure 180,000 Screens”, 18 November 2026.