Currently, there is some contention between the leading mobile OS providers, Apple and Google, and the EU Commission with regards to the Digital Markets Act (DMA) and it’s put me in a bit of a dilemma. Consumers should be able to do what they want with devices they purchase. But is there an obligation for OS developers, no matter the underlying platform (desktop, mobile, IoT, or OT), to protect the user from themselves? Let me explain.
A quick perusal of the DMA shows that it’s about ensuring fairness and competition when it comes to what they call “gatekeepers”; large digital platforms, like Amazon, Apple, Alphabet (Google), Meta, or Microsoft that provide core services like search and app stores. For Apple and Google, “gatekeepers” for mobile devices means these providers aren’t just offering an OS, but a plethora of services that have direct ties into the OS; app store, virtual assistant, search capabilities, browsers, and email to name a few. When it comes to mobile apps, Google has been transparent about fairness and interoperability as Android allows using other apps stores, side-loading of apps, and switching the default of any Google provided app to a third-party. Apple has, until rather recently, been more closed and have made many structural changes to allow third-parties access to the same functions. These changes only directly impact EU-resident of Apple (and now Japan) and that comes across as a slight to non-EU customers and developers as they should open the platforms for all customers globally, but that’s a different issue. Where things get heated is when we turn to AI and the DMA’s Article 6(7).
Yes, six seven. The meme has become something real.
There is nothing specific in the DMA with regards to artificial intelligence, only to virtual assistants and that the same level of access the native virtual assistants have should apply to any 3rd party assistant that the phone’s owner wants to use. However, because the current assistants shipped by Apple and Google are using AI, some of the arguments have become how the AI brought in through the third-party virtual assistants can access the same things that Apple and Google’s assistants’ access. This is where things can quickly go off the rails and brings us to that crossroads.
Malware on smartphones is a serious problem, and we’re not just talking about apps that are sideloaded or downloaded from 3rd party stores; both Apple and Google have had malware/spyware/trojans within apps hosted on their maintained app stores. These apps steal data, hijack identities, or possibly allow an attacker to compromise other apps and damage the user. Because of mobile OSs are designed, it’s rare when mobile malware can access the OS core and fully compromise the device. But by forcing open this door into the area where mobile virtual assistants play – interacting with the user for their input, retrieving and submitting information into any app that’s requested by the user, the changing settings within the OS, accessing the sensors, and, in the case of Google, access to the searches that the user has done through their account – the EU commissioners are playing with fire. Virtual assistants directly interact with your applications, have read/write access to system configurations, stay resident on the system, in the context of Google, have access to all other components, including search history, from your Google account, etc, and if that assistant is malicious, not only would local device data be compromised, your account data, data from other apps on the device, or even data from websites your accessed through that assistant could be compromised. But we also have to consider AI agents as virtual assistants.
AI agents act on your behalf, but they are not you. They have their own identity and when interacting with various layers on an endpoint or enterprise, they will take all necessary actions to complete their tasks and there have been occurrences where an agent took malicious actions, even when prompted not to. In the case of a smartphone, an AI-based virtual assistant could easily – either prompted by an attacker or simply by hallucination – prompt a user to take an inappropriate action and expose themselves, or their business, to compromise.
The drive for fairness by the EU commissioners and those who want control of the devices they own, and the data associated with them is fair. So long as they acknowledge that if they do something silly with those devices, like downloading a random AI-powered virtual assistant and installing it on their smartphone, they should be responsible for the consequences of their actions. But is your average user, consumer or corporate, computer and cybersecurity savvy enough to know how to avoid unsavory apps and agents. What about business leaders? And defaulting to “our endpoint security solution will pick up any malicious actions” is invalid because data from our 2026 Forrester Security Survey says only about 40% of environments are using mobile antivirus and 35% are using mobile threat defense – the equivalent of EDR for mobile. So, unless more security leaders begin deploying mobile threat defense (MTD) solutions, they’re not going to have much insight into what if their users are using their mobile devices safely. And if they’re not deploying MTD on the BYO devices that are connecting to company resources, they end up with a compromised mobile device containing a rogue virtual assistant that could pilfer data or spread malware within your organization.
Forrester clients interested in this topic should connect with me to discuss via an inquiry or guidance session.
