The Hacker News | #1 Trusted Source for Cybersecurity News — Index Page

Cybersecurity researchers have discovered a novel attack technique called TokenBreak that can be used to bypass a large language model's (LLM) safety and content moderation guardrails with just a single character change. "The TokenBreak attack targets a text classification model's tokenization strategy to induce false negatives, leaving end targets vulnerable to attacks that the implemented protection model was put in place to prevent," Kieran Evans, Kasimir Schulz, and Kenneth Yeung said in a report shared with The Hacker News. Tokenization is a fundamental step that LLMs use to break down raw text into their atomic units – i.e., tokens – which are common sequences of characters found in a set of text. To that end, the text input is converted into their numerical representation and fed to the model.  LLMs work by understanding the statistical relationships between these tokens, and produce the next token in a sequence of tokens. The output tokens are detokeni...

AI is changing everything — from how we code, to how we sell, to how we secure. But while most conversations focus on what AI can do, this one focuses on what AI can break — if you’re not paying attention. Behind every AI agent, chatbot, or automation script lies a growing number of non-human identities — API keys, service accounts, OAuth tokens — silently operating in the background. And here’s the problem: 🔐 They’re invisible 🧠 They’re powerful 🚨 They’re unsecured In traditional identity security, we protect users. With AI, we’ve quietly handed over control to software that impersonates users — often with more access, fewer guardrails, and no oversight. This isn’t theoretical. Attackers are already exploiting these identities to: Move laterally through cloud infrastructure Deploy malware via automation pipelines Exfiltrate data — without triggering a single alert Once compromised, these identities can silently unlock critical systems. You don’t get a second cha...

A novel attack technique named EchoLeak has been characterized as a "zero-click" artificial intelligence (AI) vulnerability that allows bad actors to exfiltrate sensitive data from Microsoft 365 (M365) Copilot's context sans any user interaction. The critical-rated vulnerability has been assigned the CVE identifier CVE-2025-32711 (CVSS score: 9.3). It requires no customer action and has been already addressed by Microsoft. There is no evidence that the shortcoming was exploited maliciously in the wild. "AI command injection in M365 Copilot allows an unauthorized attacker to disclose information over a network," the company said in an advisory released Wednesday. It has since been added to Microsoft's Patch Tuesday list for June 2025, taking the total number of fixed flaws to 68. Aim Security, which discovered and reported the issue, said it's an instance of a large language model (LLM) Scope Violation that paves the way for indirect prompt injecti...

Human identities management and control is pretty well done with its set of dedicated tools, frameworks, and best practices. This is a very different world when it comes to Non-human identities also referred to as machine identities. GitGuardian’s end-to-end NHI security platform is here to close the gap. Enterprises are Losing Track of Their Machine Identities Machine identities–service accounts, API keys, bots, automation, and workload identities–that now outnumber humans by up to 100:1 are in fact a massive blind spot in companies' security landscape: Without robust governance, NHIs become a prime target for attackers. Orphaned credentials, over-privileged accounts, and “zombie” secrets are proliferating—especially as organizations accelerate cloud adoption, integrate AI-powered agents, and automate their infrastructure . Secrets Sprawl: The New Attack Surface GitGuardian’s research shows that 70% of valid secrets detected in public repositories in 2022 remained active in ...

ConnectWise has disclosed that it's planning to rotate the digital code signing certificates used to sign ScreenConnect, ConnectWise Automate, and ConnectWise remote monitoring and management (RMM) executables due to security concerns. The company said it's doing so "due to concerns raised by a third-party researcher about how ScreenConnect handled certain configuration data in earlier versions." While the company did not publicly elaborate on the nature of the problem, it has shed more light in a non-public FAQ accessible only to its customers (and later shared on Reddit ) - The concern stems from ScreenConnect using the ability to store configuration data in an available area of the installer that is not signed but is part of the installer. We are using this ability to pass down configuration information for the connection (between the agent and server) such as the URL where the agent should call back without invalidating the signature. The unsigned area is u...