Skip to content
IIAIT

AI Today · 21 November 2026

Signal blocks screenshots to keep AI agents out of chats, and a House report targets Chinese open-weight models

A secure messenger restricts what operating-system AI assistants can see, two House committees question US companies' use of Chinese open-weight models, and Gemini and Canva change how much AI you get on your desktop.

  1. 01

    Signal restricts screenshots on iOS 27 and Android so built-in AI assistants cannot read your chats

    Signal said on November 20, 2026 that it now blocks screenshots inside its app on iOS 27, through a setting called Block Siri AI, to stop Apple's assistant from capturing message content. On Android it detects when an AI assistant tries to read the screen and then switches on its Screen security setting, though Signal says this does not work reliably on every Android device. It cannot block screenshots on macOS and suggests adding Signal to the excluded apps in the Siri settings of macOS 27. Signal calls the protections incomplete and asks Apple, Google and Microsoft for proper ways to keep apps out of AI agents' reach.

    What it means for you

    If your team uses Signal for sensitive conversations, leave the new protection on and, on Macs, exclude Signal from Siri. More broadly, review which apps and data your phone and desktop AI assistants can read, since screen access and file access are separate routes that one setting does not close.

  2. 02

    House committees report on US companies' use of Chinese open-weight AI models and recommend limits

    The House Homeland Security Committee and the Select Committee on the CCP published a joint report on November 20, 2026, based on a seven-month investigation of Airbnb, DoorDash and Anysphere, the developer of Cursor. It says companies adopt Chinese-developed open-weight models because they are cheap, customizable and perform well, and notes that Cursor's Composer 2 model is built on a Moonshot AI base model. Anysphere told the committees it plans to explore phasing out such models because of geopolitical risk. The 12 recommendations include an AI bill of materials and a model provenance standard, and a ban on Chinese-developed models in federal systems and federal contract work.

    What it means for you

    These are recommendations, not law, but federal contractors should expect questions about which models sit inside their AI tools. Ask your vendors, including coding assistants, which base models they use and keep a simple list of the models running in your own systems.

  3. 03

    Google adds computer use to the Gemini app for Mac, letting it work across apps and files in the background

    Google began rolling out computer use in the Gemini app for Mac on November 19, 2026. Gemini can carry out multi-step tasks across local files, desktop apps and the web, such as pulling details from PDFs or building a Google Slides deck from dashboard screenshots, in a separate background window while you keep using your screen. You choose which apps it can access, watch a live preview, can pause at any time and approve key actions. It is available in English in the US for personal accounts aged 18 and over on Google AI Plus, Pro or Ultra plans, on Apple Silicon Macs with macOS Sequoia 15.0 or later.

    What it means for you

    This is for personal accounts, not business accounts, so check your company's policy before pointing it at work files. If you try it, grant access to as few apps and folders as possible and review each action Gemini asks to approve.

  4. 04

    Canva lets Free and Pro users switch off most generative AI features and see less AI-made content

    Canva introduced two settings for Free and Pro accounts on November 19, 2026. A switch under Settings, Permissions turns off generative tools such as Magic Write, Magic Design, Magic Expand, Magic Edit, Dream Lab, Canva Code and AI Shapes, while Background Remover and Magic Grab stay available because they edit existing images. A second setting under Your account, AI personalization shows fewer AI-generated templates and elements in the library, but does not remove them entirely. Admins of Teams, Business, Enterprise and Education accounts already had similar controls.

    What it means for you

    If clients or your brand rules limit AI-generated material, staff on individual Canva plans can now turn the generative tools off themselves. On team plans, check that the admin AI permissions match your policy, and remember the library filter reduces AI content rather than excluding it.