Researchers from Microsoft have introduced a new way to spot malicious modifications in AI models without knowing what to look for. This method targets large language models (LLMs) that are often shared or reused, which can hide secret triggers. These hidden backdoors, called “sleeper agents,” can remain inactive during normal testing but activate harmful behaviors
Loading
All posts tagged in AI Safety
AI technology has been advancing quickly over the past year, but the ways we test and manage its risks aren’t keeping pace. A new report, the International AI Safety Report 2026, highlights these gaps. It brings together insights from more than 100 experts across over 30 countries, showing how testing methods are struggling to reflect
Read More
183
Mozilla is making a bold move with its upcoming Firefox 148 update, launching on February 24th. The browser will include a new “AI controls” section in settings, giving users a simple way to block all generative AI features. This is a notable shift in the tech world, where AI integration often feels unavoidable and pushed
Read More147
Many engineering teams are eager to adopt AI tools, but they often struggle to understand how mature their AI practices are. To help address this, Coder has launched a free AI Maturity Self-Assessment. This tool is designed to give organizations a clear picture of their current AI capabilities, so they can better plan their next
Read More150
Apiiro has introduced a new tool called Guardian Agent, an AI-powered safeguard designed to prevent the creation of insecure or non-compliant code by AI coding assistants. The company announced that Guardian Agent is now available in a private preview, aiming to change how software security is handled in AI-driven development environments. What is Guardian Agent
Read More164
Apple has recently updated its Platform Security guide, highlighting a major upgrade for M5 Macs. The new security features promise to significantly boost protection against advanced cyber threats. This move underscores Apple’s commitment to keeping its devices secure in an ever-evolving digital landscape. Memory Integrity Enforcement Brings Stronger Defense The highlight of the update is
Read More155
It’s Data Privacy Week, a good reminder of how our personal information is often at risk. While recent years have seen setbacks in privacy protections, Apple has made a positive move this week. The company announced that it is blocking mobile carriers from tracking your precise location through your network signal. This is a step
Read More196
WhatsApp has rolled out a set of new security features called “Strict Account Settings” to help keep users safer. These features need to be activated manually and are designed to block suspicious files and attachments from unknown senders, prevent link previews that could lead to scams, and silence calls from unknown numbers. The goal is
Read More176
The U.S. Department of Energy (DoE) is quietly rewriting nuclear safety and security standards as part of a major push to develop new types of nuclear reactors. This move is tied to a program aiming to launch three experimental commercial nuclear reactors by July 4. A recent report from NPR reveals that these changes are
Read More187
A new report from Deloitte warns that companies are rolling out AI agents faster than their safety protocols can keep up. This rapid deployment raises serious concerns about security, data privacy, and accountability. Many businesses are moving from testing phases directly into full production without implementing enough safeguards. Rapid Adoption Challenges Traditional Risk Controls According
Read More204
