LIVE · cybersecurity feed
Live wire
How Cloudflare WAF helps protect against phpAfterShell and other critical PHP flawsCVE-2026-114938 · A Technical Guide to Mitigate Java4Shell with TruRisk EliminateCVE-2026-104995 · Dontdream BP Profile Search Vulnerability Listed As Exploited In VulnCheck KEVCVE-2026-54836 · YMC Filter SQL Injection Vulnerability Added to VulnCheck KEVCVE-2026-7542 · Slider Revolution Sensitive Information Exposure Added To VulnCheck KEVCVE-2024-8517 · SPIP External Control Vulnerability CVE-2024-8517 Added To VulnCheck KEVSean Cairncross out as national cyber directorVMware patches vCenter zero-day exploited in attacks since SeptemberASUS warns of critical hard-coded key flaw in Control Center ExpressCVE-2026-123890 · CVE-2026-123890: Critical Authentication Bypass in VMware vCenter
ai

Signal slams door on prying AI agents to protect privacy

Signal has implemented new restrictions within its mobile applications to prevent AI agents from accessing private user communications. The encrypted messaging platform now blocks automated tools—specifically Google Gemini on Android and Siri on iOS—from capturing screenshots of message content. This update is designed to stop these operating system-level agents from using optical character…

ZeroDay News ·

Source: The Register — Security

Signal has implemented new restrictions within its mobile applications to prevent AI agents from accessing private user communications. The encrypted messaging platform now blocks automated tools—specifically Google Gemini on Android and Siri on iOS—from capturing screenshots of message content. This update is designed to stop these operating system-level agents from using optical character recognition to extract and process text from private conversations, which could otherwise lead to that data being uploaded to external servers.

Signal CEO Meredith Whittaker confirmed that these measures were prompted by the increasing integration of agentic capabilities into mobile operating systems. By scanning the information displayed on a user's screen, these AI agents can ingest private data to facilitate their own functionality. Signal has enabled these protections by default, though users retain the ability to disable them in the settings menu if they wish to allow screenshots, with the company explicitly warning that doing so exposes their data to potential AI harvesting.

The company acknowledged that screenshot blocking is an incomplete defense, as AI agents may still find alternative methods to access application data through the underlying device file system. Whittaker stated that Signal and other application developers currently lack robust, standardized tools from major platform providers like Apple, Google, and Microsoft to effectively restrict AI agent access. According to Signal, the current technical landscape forces developers to choose between protecting user privacy and maintaining core application functionality.

These limitations are particularly evident on desktop platforms. Signal noted that it is currently unable to implement similar protections on macOS due to the way the operating system handles data access and application permissions. While Apple is introducing updates in macOS 27.3—currently in public beta—to provide more granular controls over Device Control, Data Access, and Full Disk Access APIs, these changes are primarily focused on managing third-party access rather than restricting Apple’s own integrated services.

Apple has stated that its upcoming changes to Full Disk Access are intended to ensure users are better informed about the risks associated with granting broad permissions to software. The company emphasized that as AI agents become more autonomous, the necessity for clear user consent and understanding regarding data exposure will increase. However, the broader industry remains divided on how to balance the utility of AI agents with the need for data security.

The move by Signal reflects a growing trend of organizations erecting barriers against autonomous AI agents. Various companies across retail, finance, and e-commerce have begun implementing similar restrictions to protect their platforms and user data. For instance, Amazon has blocked shopping agents from its site, while entities such as Walmart, United Airlines, and several major financial institutions have established their own policies to govern or outright ban interactions with unauthorized AI bots.

This shift in strategy highlights a fundamental tension regarding who these AI agents ultimately serve. While developers of agentic software argue that access to digital data is necessary for personalization and context, many service providers are increasingly viewing these agents as threats to privacy and established business relationships. As a result, a specialized industry focused on bot detection and mitigation is expanding, as companies seek to regain control over how their data is accessed and utilized by third-party automated systems.

ai
ShareXLinkedInWhatsAppFacebook

More News

view all →
ai

US companies’ use of Chinese open weight AI draws rebuke from House chairmen

A joint investigation by the House Homeland Security Committee and the Select Committee on the Chinese Communist Party has issued a stern warning regarding the integration of Chinese-developed open-weight artificial intelligence models into American corporate infrastructure. The 50-page report identifies a growing trend among both Fortune 500 companies and Silicon Valley startups to utilize…

patch

PHP 8.6: Things are gonna get a lot more strict around here

The PHP development team has officially released PHP 8.6, marking the latest iteration of the widely used server-side scripting language. This release arrives following a turbulent period for the platform, which has faced a surge of security vulnerabilities since October. These flaws, which allowed unauthenticated attackers to trigger remote code execution (RCE) or crash PHP workers,…

security

4 Best XRP Lending Platforms Compared by Security, LTV, and Speed

A recent comparative analysis has evaluated four prominent platforms facilitating XRP lending, focusing on the trade-offs between security protocols, loan-to-value (LTV) ratios, and transaction speeds. The review examines CoinRabbit, Arch, Nexo, and Coinbase, providing a technical overview of how these services manage collateralized digital asset lending.

CVE-2026-23871

CVE-2026-23871: WhatsApp Bug Helped Scammers Make Messages Look Legitimate

WhatsApp has released security updates addressing three vulnerabilities, one of which was actively exploited in a scam campaign designed to increase the perceived legitimacy of fraudulent messages. The company identified this specific flaw, tracked as CVE-2026-23871, while investigating and disrupting an ongoing fraud operation.

cloud

Dialing for Data: Inside the Com’s High-Profile Heists

Security researchers have identified a persistent and evolving campaign of vishing-led cyberattacks targeting data-intensive industries across the United States, the United Kingdom, and Australia. Since August 2026, a network of threat actors associated with the Com collective has conducted a series of data theft and extortion operations. These attacks primarily target the financial,…

phishing

Visa, Mastercard theft included info useful for phishing, but not draining bank accounts, say researchers

Visa, Mastercard, and JPMorgan Chase have confirmed they are investigating data breaches involving third-party service providers. The incidents came to light after a ransomware and extortion group known as n0n claimed responsibility for the thefts and began publishing samples of the exfiltrated data to pressure the financial institutions.