Follow on Telegram

tech/security

Norway Issues Guidance on AI Use in Education

Norway's National Security Authority (NSM) has issued guidance recommending that AI tools, including large language models, not be used in educational contexts. The advice stems from concerns regarding data privacy and security, particularly the potential for sensitive personal information to be processed or exposed. The NSM suggests that

Technology News

US Government Forces Anthropic to Pull New Models Amid Security Concerns

The US government has ordered Anthropic to withdraw its two newest AI models, Fable 5 and Mythos 5, citing national security concerns. The action came after Amazon researchers reportedly discovered a method to circumvent Fable 5's built-in safety guardrails. The decision has drawn criticism from cybersecurity researchers, who

US Government Forces Anthropic to Pull Fable 5 Models Over Security Concerns

The US government has ordered Anthropic to pull its newest AI models, Fable 5 and Mythos 5, from public availability. The action was taken citing national security concerns after Amazon researchers reportedly discovered ways to circumvent Fable 5's built-in safety guardrails. The decision has drawn sharp criticism from

Technology News

Anthropic Export Controls Strain US Relations With Technology Partners

US export controls targeting Anthropic's advanced AI technology are reportedly straining relationships with international technology partners. The controls, designed to prevent sensitive AI capabilities from reaching foreign adversaries, have raised concerns among allied nations about access to cutting-edge AI tools and collaborative research opportunities. Industry observers note that

Microsoft Warns of Remote Code Execution Risk in Web-Enabled AI Agents

Microsoft has raised security concerns about web-enabled AI agents, warning that they can potentially enable remote code execution (RCE) at the host level. RCE vulnerabilities are particularly serious because they allow attackers to execute arbitrary code on a target system, potentially gaining full control over the affected machine. The warning

Security Researchers Identify New Prompt Injection Vulnerability in Microsoft 365 Copilot

Security researchers have identified a new vulnerability in Microsoft 365 Copilot that could expose organizational data through prompt injection attacks. The vulnerability, referred to as "SearchLeak," specifically targets the search integration within Microsoft's AI-powered productivity suite. Prompt injection attacks involve maliciously crafted inputs that trick AI

SK Telecom Caught in Anthropic's Mythos Export Controls Dispute

SK Telecom Involved in Anthropic Export Controls Controversy SK Telecom has been named as the Korean carrier at the center of a controversy involving Anthropic's Mythos AI model and U.S. export controls. The situation escalated when access was revoked just days before the White House took action