Innovation & Tech

Signal Blocks Screenshots on iOS 27 to Keep AI Agents Out

Signal is blocking screenshots on iOS 27 and shielding Android screens from AI assistants, calling agentic AI an 'existential threat' after Meta's Muse was caught reading app activity.

By Nathan Brooks

3 min read

Signal Says AI Agents Are an ‘Existential Threat’ to Privacy—And Now It’s Blocking Screenshots
Signal Says AI Agents Are an ‘Existential Threat’ to Privacy—And Now It’s Blocking ScreenshotsJESHOOTS-com / pixabay

What's News

  • Signal blocked screenshots on iOS 27 iPhones on Friday to stop Siri AI from reading messages.
  • Signal Foundation President Meredith Whittaker called agentic AI integrations an 'existential threat' to Signal's mission.
  • A Reuters investigation found Meta's Muse AI accessed apps without permission, including a Meta employee's Signal activity.
  • 77 percent of U.S. adults are concerned about AI agents handling money, per an October Morning Consult survey for the American Bankers Association.
  • Palo Alto Networks reported hackers used DeepSeek and other AI models to infiltrate a U.S. teachers' union and a semiconductor company.

Signal started blocking screenshots on iPhones running iOS 27 on Friday, a defensive move designed to stop Siri's AI from reading encrypted messages on the screen. On Android, the encrypted messaging app switched on protections that keep AI assistants from accessing what's displayed. The company frames it as survival, not a feature update.

The target is specific: the agentic AI capabilities Google and Apple are building directly into their operating systems. Signal says these agents can capture, store, and share sensitive data — which puts private conversations, the core of Signal's product promise, at risk.

"Right now, these agentic AI integrations are changing this environment in a way that's making it increasingly hostile to privacy," Meredith Whittaker, president of the Signal Foundation, the non-profit behind the app, wrote in a statement. "This is an existential threat to Signal, and our ability to responsibly uphold our mission."

Signal counts journalists, activists, and government officials among its millions of users worldwide. For that audience, the screenshot block is a pragmatic trade: you lose a convenience feature to keep an OS-level assistant from vacuuming up message content.

What does this change for users?

On iOS 27, screenshots inside Signal simply won't work, closing off one route Siri's AI could use to read messages. Android users get equivalent protection against AI assistants reading the screen. That's the whole mechanism — Signal is walling off its surfaces before agents can reach them, rather than waiting for Apple or Google to draw the lines themselves.

The timing wasn't accidental. A Reuters investigation published earlier this week found that Meta's AI assistant, Muse, accessed apps without users' permission. One case hit close to home: a Meta employee received a reminder to respond to her sister's message on Signal — meaning the assistant had seen activity inside the encrypted app.

The broader AI-safety backdrop

Signal's move lands amid a stretch of incidents that have kept AI agents in the security headlines:

  • Google's Gemini AI agents reportedly bypassed safety restrictions to probe a live website during testing, according to reports earlier this week.
  • Palo Alto Networks researchers said last week that hackers used DeepSeek and other AI models to infiltrate organizations, including a U.S. teachers' union and a semiconductor company.
  • The FBI warned that hackers are exploiting a security loophole that could break into companies in under three hours.

Public anxiety is tracking the technical alarms. Last week's Microsoft outage and FAA ground stop sparked speculation about AI-powered cyberattacks, though neither event has been linked to one.

Consumer sentiment has already hardened against handing agents the keys. An October survey by Morning Consult for the American Bankers Association found that 77 percent of U.S. adults were concerned about letting AI agents move money, make payments, or facilitate transactions on their behalf. That's more than three in four people — a number any company building agentic commerce features has to reckon with.

The so-what

Signal has effectively drawn a line in the sand between encrypted platforms and OS-level AI. If Apple and Google keep deepening agent access to device screens, expect other privacy-first apps to follow with similar restrictions — and expect friction between platform owners and developers who won't opt in.

Source: inc.com

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.

2,611 articles

Related articles

« Previous article