# Signal slams door on prying AI agents to protect privacy

> Source: <https://www.machinebrief.com/news/signal-slams-door-on-prying-ai-agents-to-protect-privacy-r23p>
> Published: 2026-11-20 20:42:04+00:00

# Signal slams door on prying AI agents to protect privacy

[The Register](https://www.theregister.com)

Apple, Google, and Microsoft called out for failing to provide developers with adequate AI agent controls

Signal, the encrypted messaging app, has barred Google [Gemini](https://www.machinebrief.com/glossary/gemini), Apple Siri, and other AI agents from accessing messages unless the user has explicitly granted permission, based on privacy concerns.

Citing the screenshot blocking implemented last year to defend against Microsoft Recall, Signal CEO Meredith Whittaker said similar measures have become necessary to prevent agentic software available in iOS and Android from gaining access to communication that's intended to be private.

"Google and Apple are integrating agentic capabilities like [Gemini Intelligence](https://support.google.com/gemini/answer/15850607?hl=en) with Google Gemini and [Onscreen Awareness](https://www.apple.com/newsroom/2026/09/siri-ai-a-profoundly-more-capable-and-personal-assistant-is-here/) with Siri AI deep into their operating systems," wrote Whittaker in a [blog post](https://signal.org/blog/protecting-your-privacy/).

"In order to gain access to information about you, these OS AI agents 'scan' the information that’s on your screen. In doing so, your Signal data is captured, then made available to these agentic systems in order to allow them to answer your questions and do your bidding."

The result is that private messages ingested by these agentic systems may end up on someone else's server where they're no longer private.

Updates to Signal's iOS and Android apps now include controls that by default block Siri on iOS and Gemini and other agents on Android from taking screenshots, which AI agents process using optical character recognition to extract text.

Whittaker says that those who wish to take screenshots of their messages can disable this protection via Signal's settings menu, but warns doing so risks exposing your Signal content (and that sent by your contacts) to AI agents.

Screenshot blocking, Whittaker concedes, is only a partial solution since AI agents may have other ways to access application data through the device file system.

"We want to be very clear: it is a serious problem that Signal – and application developers beyond – lack robust tools to prevent AI agents from undermining application-level privacy," she said.

"Even where we are able to detect and block AI data access, Google, Apple, Microsoft, and others do not provide ways for developers to easily restrict such AI data access without making drastic compromises to the core user experience. This is also the reason we are currently unable to implement these protections on macOS."

Apple is taking some action to implement protection from AI agents on macOS, though it appears to be doing so largely to defend against third-party access to data rather than to limit its own Siri service. Siri can be turned off separately.

[With macOS 27.3](https://developer.apple.com/news/?id=oqng0zjk), currently in public beta testing, Apple's desktop operating system will get more granular controls on two APIs: Device Control and Data Access, and Full Disk Access. The result will be more permission prompts for users of macOS apps that seek access to user data, and the inconvenience of applying for an entitlement from Apple to obtain Full Disk Access for backup software and assistive technologies that use Accessibility APIs.

"As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially," Apple [said](https://developer.apple.com/news/?id=p6zjojqw) when it introduced the Full Disk Access change. "We are committed to ensuring users clearly understand these risks before granting such access, so they can make informed decisions about their own data and privacy."

Apple is also [reportedly](https://www.bloomberg.com/news/articles/2026-11-19/apple-is-working-on-ai-agent-for-device-control-apps-browsing-and-coding) working on its own [AI agent](https://www.machinebrief.com/glossary/ai-agent) in a bid to compete with [Anthropic](https://www.machinebrief.com/glossary/anthropic) and [OpenAI](https://www.machinebrief.com/glossary/openai). It remains to be seen whether Apple will use its operating system ownership as competitive leverage against AI firms that threaten its gatekeeping power, [as it has done with web browsers](https://www.theregister.com/software/2026/06/17/apples-webkit-performance-tax-leaves-ios-browsers-stuck-in-the-slow-lane-says-microsoft/5257384).

### First the flood, then the dams

While Apple casts itself as an ally that's helping to defend device customers against prying AI agents, Signal's decision to block Siri shows that AI agents have complicated trust decisions and the competitive landscape. AI agents are supposed to serve the user but when they spirit data away to the cloud and come between customers and merchants, they wear out their welcome.

AI agents also threaten the business relationship between consumers and merchants.

Amazon has blocked shopping agents like [Meta Muse](https://www.geekwire.com/2026/amazon-blocks-metas-muse-ai-assistant-in-new-standoff-over-agentic-shopping/) from its e-commerce site (though [not AWS Marketplace](https://aws.amazon.com/blogs/awsmarketplace/aws-marketplace-helps-humans-and-agents-buy-deploy-and-transact/)) while Walmart has [invited third-party agents](https://corporate.walmart.com/news/2026/10/Making-It-Possible-for-Agents-to-Shop-on-Behalf-of-Customers) to interact through its AI Gateway. Anthropic has blocked subscription customers [from using third-party harnesses](https://www.theregister.com/software/2026/02/20/anthropic-clarifies-ban-on-third-party-tool-access-to-claude/5014546). United Airlines has [banned](https://www.united.com/en/us/fly/help-center/digital-trust-center/artificial-intelligence/third-party-ai-agents.html) transactions from unauthorized AI bots. Similar barriers or limitations related to third-party AI agents have been erected by Target, Home Depot, Etsy, eBay, JPMorgan Chase, Charles Schwab, Citigroup, and Wells Fargo, among others.

Meanwhile, there's a burgeoning industry in bot detection and mitigation, led by the likes of Cloudflare, Akamai, and DataDome, to name a few.

The web browser is sometimes referred to as a user agent, based on the idealistic notion that it serves its user. That vision has been muddied by online advertising and data collection, and default settings that often serve the interests of browser makers instead of promoting privacy.

Now AI agents face the same scrutiny as people wonder whether their agents really serve them or someone else. After a year or two of agent vendors arguing for access to all digital data for the sake of personalization and context, the backlash has arrived. ®

Get AI news in your inbox

Daily digest of what matters in AI.

## Key Terms Explained

[AI Agent](https://www.machinebrief.com/glossary/ai-agent)

An autonomous AI system that can perceive its environment, make decisions, and take actions to achieve goals.

[Anthropic](https://www.machinebrief.com/glossary/anthropic)

An AI safety company founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei.

[Gemini](https://www.machinebrief.com/glossary/gemini)

Google's flagship multimodal AI model family, developed by Google DeepMind.

[OpenAI](https://www.machinebrief.com/glossary/openai)

The AI company behind ChatGPT, GPT-4, DALL-E, and Whisper.
