Signal has rolled out new privacy protections on iOS 27 and Android to prevent system-level artificial intelligence assistants from reading private conversations.
The messaging application announced the security updates on Friday, November 20, in a blog post written by President Meredith Whittaker. Whittaker said that the rise of AI agents integrated directly into operating systems creates an environment increasingly hostile to user confidentiality, calling it an existential threat to Signal and its end-to-end encryption guarantees.
Encryption secures messages during transit across networks. However, that protection ceases once plaintext data renders visually on a device display. Features like Gemini Intelligence on Google Android and Onscreen Awareness under iOS 27 scan on-screen data to parse user prompts, transmitting captured text to local or remote servers outside of Signal’s architectural control, according to Frandroid.
To mitigate this exposure, Signal 8.37 on iOS 27 introduces a new feature called “Block Siri AI.” The setting switches on by default, preventing application screenshots and masking conversation previews inside the App Switcher to stop background OS processes from harvesting visible text strings, according to Iphonesoft.

“These protections are meant to stop operating system agents from capturing and storing Signal data,” Meredith Whittaker told ai.info on November 20.
Users who prefer retaining manual image capture capabilities can disable the defense via Signal > Réglages > Confidentialité > Block Siri AI. The menu provides options to keep the shield active, disable it completely, or suspend enforcement temporarily for a ten-minute window.
Protection strategies differ between Apple and Android devices because their operating systems handle security updates and hardware integration in fundamentally different ways.
On Android devices running version 8.35.1 or newer, Signal automatically trips its screen security layer whenever an active AI process attempts interface interaction. The application avoids blanket screenshot blocks on select Android builds where hardware fragmentation prevents reliable capture detection, prioritizing process identification over outright interception.
Signal acknowledges that intercepting screen captures provides an incomplete barrier against aggressive ambient intelligence harvesting.
Blocking image grabs does not stop background frameworks from reading cached data stored directly within file system directories. Whittaker urged major platform developers to supply standardized, documented APIs allowing software engineers to programmatically exclude sensitive applications from ambient processing routines without breaking core utility.
Meredith Whittaker, who wrote the blog post announcing the new protections, spent 13 years at Google where she founded the Open Research group and co-founded M-Lab, a global network measurement platform.
In 2018, she was a core organizer of the Google Walkouts involving over 20,000 employees protesting the company’s handling of sexual misconduct allegations and its involvement with Project Maven, a U.S. military drone program. She resigned from Google in July 2019 and later co-founded the AI Now Institute at New York University to study the social implications of artificial intelligence.
Because these system hooks remain inadequately documented, macOS users must manually add Signal to the excluded applications list under Siri system settings. The messaging platform continues calling for explicit developer controls covering screen reading permissions, notification scraping vectors, and local filesystem APIs.
Keep reading
- GTA 6 Xbox Series X 60 FPS Workaround Found via Refresh Rate Trick but Gameplay Speeds Up
- Attribution of Summer Total Precipitation Changes in China: Dominant Role of Circulation Pattern Frequency Driven by Greenhouse Gas Forcing
- Signal Blocks Screenshots on iPhone, Android to Thwart AI Screen Access (time.news)
- Fifth Circuit Blocks Pregnant Workers Fairness Act in Three States (world-today-journal.com)