Signal Blocks AI Screen-Reading, Plus 4 Stories
Update, November 20, 2026: We clarified how Signal’s screenshot protections differ on iOS and Android and added Broadcom’s patch guidance for vCenter 8.0 and 7.0.
Signal is the lead today because it exposes a privacy problem that end-to-end encryption can’t solve by itself. If an AI agent can see your screen after Signal has decrypted a message, the cryptography did its job and the message can still leave the app. The rest of the brief follows the same practical thread: where access really sits, who gets to use it, and what happens when a supposedly narrow permission turns out to be broad.
Signal blocks local screenshots to keep AI agents out of chats
Signal says it is blocking in-app screenshots on iOS 27. On supported Android devices, Signal turns on Screen Security when it detects an AI assistant trying to read the screen. The immediate trigger is agentic AI: software that doesn’t merely answer a question, but can use accessibility controls, screenshot tools, notifications, and other operating-system permissions to move through apps on a user’s behalf.
That distinction matters. Signal’s end-to-end encryption protects a message while it travels between devices and limits what Signal can read on its own servers. Once the recipient’s phone or computer decrypts the message and draws it on the screen, an AI agent with permission to see that screen may be able to copy it into another service. The attacker doesn’t need to break Signal’s encryption. The operating system politely hands over the finished product.
Reuters reported that Meta’s Muse agent used broad permissions to read Signal messages on macOS and Android, including a combination of accessibility and screenshot tools on the Mac. Meta says the agent accessed other apps only through permissions users explicitly granted. That’s true as far as it goes. It also demonstrates why “the user clicked Allow” is a weak privacy model when one approval can be reused for a purpose the user wasn’t thinking about.
Signal calls its protection incomplete, and it is. Behavior varies by operating system. Signal says it can’t reliably provide the same protection on macOS, and its screen-security documentation is clear that blocking a screenshot on your device doesn’t stop the person at the other end from capturing the conversation. Accessibility tools can also be essential for people with disabilities, so simply denying every screen-reading capability is not a serious universal answer.
Update Signal, leave screen security enabled unless you have a specific reason not to, and review which AI assistants have screenshot, accessibility, notification, and file access. For business conversations, remember that “encrypted” describes the connection and the app’s design. It doesn’t bless every tool running beside the app. This is the same permission problem we’ve been following in AI agents and other trusted tools, now arriving in a place people reasonably thought was private.
Turning on AI is the easy part
Deciding what an AI tool should be allowed to see, who should use it, what work it should perform, and what happens when it gets something wrong is the more interesting problem. Raymond Tec helps businesses connect and automate the tools they actually use without treating every new feature like a button that obviously needs to be switched on.
Bank and card marketing data escaped through vendors
JPMorgan Chase, Visa, and Mastercard have confirmed separate third-party incidents involving customer information. Chase says outside vendors may have exposed customer names and email addresses from marketing data. Visa says its incident involved a service provider in Asia Pacific and did not affect Visa systems or payment-card numbers. Mastercard says a provider exposed a limited number of email addresses belonging to customers enrolled in a benefit program, without affecting its payment network.
An extortion group has posted larger claims and sample records that reportedly include some phone numbers, postal addresses, employee addresses, and marketing interactions. Reuters reviewed some of that material but couldn’t independently verify its authenticity. That uncertainty belongs in the story. A criminal’s sales pitch isn’t a breach notification.
The confirmed data still matters because a message that knows your bank, email address, location, or card benefit is easier to mistake for the real thing. Watch for direct notices, but don’t click a link merely because the sender says it relates to this breach. Open the official banking app or call the number printed on the card. There is no reported payment-card exposure that justifies replacing a card solely on today’s information.
For businesses, this is another reminder that “we don’t store it” can mean “we sent it to three marketing vendors.” The useful inventory includes copies held by service providers, retention periods, and who can export them, not just the database in your own office. Trusted vendors are still part of the attack surface.
EU borrowers gain rights over automated credit decisions
New European Union consumer-credit rules apply today, expanding protections to loans of up to €100,000 and bringing many buy-now-pay-later agreements into the same framework. Lenders must provide standardized cost information, show the annual percentage rate and total repayment, assess whether the borrower can repay sustainably, and generally allow a 14-day withdrawal period.
The technology point is easy to miss inside the finance rules. When a lender uses automated processing of personal data to assess creditworthiness, the consumer can request human intervention and a review. The European Commission’s summary also says advertising must include clearer standard information and warnings, while misleading practices and forced product bundles are generally prohibited.
A human review is not a promise that the answer changes. It is a chance to challenge bad data, explain circumstances an automated score missed, and require somebody to own the decision. EU consumers should use that right when a credit decision seems inexplicable. Ecommerce businesses offering deferred payment should verify that their provider’s disclosures, cancellation process, and review path actually match the new rules. A logo at checkout is not a compliance plan.
The rules around technology matter too
Platforms, privacy, speech, competition, surveillance, copyright, and regulation increasingly determine what technology companies can build and what the rest of us have to live with. Browse more Raymond Tec News for practical coverage of technology policy and digital rights.
VMware vCenter has a critical login bypass
Broadcom has patched a critical VMware vCenter authentication bypass, CVE-2026-123890, in the VMware Directory Service. An attacker who can reach the vCenter management service over the network may be able to bypass authentication and gain unauthorized access. That is the console used to manage virtual machines, hosts, permissions, and much of the infrastructure living underneath ordinary business applications.
Broadcom’s updated VMSA-2026-0009.1 advisory lists 9.1.1.0200 for vCenter 9.1, 9.0.2.0200 for 9.0, and 8.0 U3m for 8.0. For vCenter 7.0, Broadcom directs customers with extended support to contact Broadcom Support. Check the advisory’s response matrix for your exact product and version. Broadcom lists no workaround for the vCenter login bypass. It also fixes separate SDDC Manager flaws and a VMware Tools issue on Windows. The vulnerabilities were privately reported; Broadcom’s advisory doesn’t say they are being exploited in the wild.
If you run VMware, confirm the exact product and version, restrict management interfaces to the administrators and networks that need them, apply the vendor update, and verify the running version afterward. Network restriction reduces who can attempt the flaw. It doesn’t repair the flaw, which is why “the console isn’t public” is a layer of protection, not permission to skip the patch.
Business IT goes well beyond the website
Your business also depends on workstations, cloud accounts, browsers, Wi-Fi, remote access, collaboration tools, and all the other technology that quietly becomes infrastructure. Raymond Tec works across that whole stack, whether the problem lives on a server, on a desk, or somewhere in between.
Smart glasses fail some very ordinary safety tests
The Dutch Data Protection Authority tested six types of AI-enabled camera glasses and advised consumers not to buy them for now. According to Reuters’ report on the regulator’s findings, several devices had security weaknesses that could let someone nearby connect without permission, access recordings, or operate the camera. The regulator did not name which products failed which tests.
The failures weren’t limited to quiet recording. In a simulated medical emergency, one pair reportedly told the wearer that an ambulance was coming when none had been called. Another gave inappropriate advice about an adult dating a minor. Those examples don’t prove every smart-glasses model is unsafe, but they do puncture the idea that this is just a normal pair of glasses with a clever assistant tucked inside.
If you’re considering them, ask where recordings go, whether the camera indicator can be obscured, how nearby connections are authenticated, and what the device does when its AI is wrong. Workplaces and venues need rules for bystander recording before the glasses arrive, not after the first uncomfortable meeting. No immediate setting fixes the wider social problem: people around the wearer didn’t agree to become input.
Still in a reading mood? The Raymond Tec News archive covers security, AI, small-business technology, policy, and the places technology collides with ordinary life — without requiring a computer-science degree to get through it.
Sources / Further Reading
- Reuters: Signal’s new AI-agent screen protections and platform limits
- Signal: Screen Security settings and limitations
- Signal: Why screen capture threatens private messaging
- Reuters: Chase, Visa, and Mastercard vendor incidents
- European Commission: Consumer credit protections applying November 20
- European Commission: Rights under the Consumer Credit Directive
- Broadcom: VMSA-2026-0009.1 security advisory
- Reuters: Dutch regulator’s smart-glasses testing
- European Data Protection Board: Statement on smart glasses
Photo by Zulfugar Karimov on Pexels.
