SecurityAWS AgentCore Flaw Let One Prompt Hijack All Agents
Researchers say a single chat prompt could extract cloud credentials and let an attacker take over every AgentCore agent in an AWS account and region.
> Search
30 results for “injection”
Security
CloudAWS Agent Flaws Expose Control Gap
Repeated AWS agent security fixes that later resurfaced show the tension between agent autonomy and isolation, researchers say.
AI & MLCopilot CLI's Model Lottery Problem
Adversa AI says encrypted 'zombie instructions' on web pages can extract secrets through GitHub Copilot CLI, depending on which model handles the session.
SecurityFortra BoKS Patches Fix Critical Auth Bypass
Fortra fixed eight bugs in BoKS, including a 9.9-severity auth bypass and a 9.1 command injection, but reported no known exploitation.
SecurityAI Agents' Hack Attempts on Gov Sites
Research lab Transluce reports autonomous AI agents tried SQL injection and other probes against U.S. and Canadian government websites.
SecurityZimbra Flaw Exploited Before Disclosure
Microsoft says attackers probed and exploited a Zimbra command injection flaw in the window between patch release and public disclosure.
Cyber CrimeZimbra Bug Exploited for Mailbox Theft
Microsoft says attackers chained a patched Zimbra command-injection flaw into web shells, credential theft and cloud exfiltration.
SecurityWatchGuard Patches 15 Fireware OS Flaws
WatchGuard fixed a critical Fireware OS code injection flaw and 13 high-severity bugs, plus three Access Point vulnerabilities disclosed a day earlier.
AI & MLAI Watermarks Can Weaken Model Safety
New research finding watermarking alters model behavior, including refusal of harmful requests and tool calling.
SecurityIn-Memory PHP Shell Sidesteps Disk Scans
F5 BIG-IP APM malware hides a PHP web shell purely in memory, evading file checks, Sophos analysis shows.
AI & MLHidden Prompt Injection Threats Rise
Bowbridge warns hidden prompt injections can hijack AI agents, urging content scanning defenses.
SecuritySangoma Switchvox flaw exploited in active attacks
CVE-2026-9586, an unauthenticated SQL injection in Switchvox, is under active exploitation, Horizon3 reports.
SecurityMalware Lures LLMs Toward Forbidden Prompts
Russia-aligned UAC-0099 embeds nuclear-weapon-style prompts in a VBS script to misdirect AI-assisted malware analysis, ESET says.
SecurityServiceNow Tackles Maximum-Severity Flaws
Patches cover three critical code injection bugs and a sandbox escape in the Now Platform.
AI & MLClaude Code Hijack via Website Summary
Researchers show Anthropic's Claude Code can be tricked into running malicious code by summarizing a website.
SecurityFirst 24 Hours: Responding to an AI Agent Incident
What to do when an autonomous agent acts beyond its bounds, hour by hour.
SecurityHidden HTML Hijacks AI Email Summarizers
Forcepoint shows invisible text can silently change what an AI assistant reads in your email.
SecurityMemory Poisoning: The New AI Attack Frontier
InjecMEM lets attackers plant persistent instructions in AI agents' memory with a single prompt.
SecurityGitLab Flaw Exploited Within Days of Patch
GitLab's CVE-2026-19478 is under active exploitation, days after disclosure, urging immediate patching.
SecurityGrok tricked into leaking user data
A researcher found that encrypting malicious instructions lets Grok exfiltrate user chats and personal details.
SecuritySnowflake CI/CD Flaw Opens Door to Issue-Based Injection
A GitHub Actions workflow flaw in Snowflake's connector repo allowed crafted issues to execute commands with exposed Jira credentials.
SecurityMCP Servers: A New Secret-Leak Vector
Model Context Protocol servers can expose enterprise secrets via plaintext configs, over-permissioning, and prompt injection, often undetected.
AI & MLOpenAI's Computer History logs your keystrokes
OpenAI's Computer History feature records clicks and typing, storing them unencrypted and expanding prompt injection risks.
SecuritySAP's Critical Patches Close Zero-Day-Style Gaps
SAP's August 2026 patch batch addresses 28 flaws, including a 10/10 severity bug in Commerce Cloud that could allow attackers to bypass authentication and execute code.
AI & MLHow MCP Splits Can Leak Secrets
Malicious MCP servers can split instructions to make AI coding agents exfiltrate secrets, ASSET reports.
SecurityFlaw in Progress LoadMaster exploited in attacks
CISA warns of active exploitation of critical command injection flaw in Progress Kemp LoadMaster; urges patching.
SecurityMetabase Zero-Day Exploited Before Patch
Critical SQL injection flaw in Metabase allowed zero-day attacks; urgent patches released for self-hosted users.
SecurityAtlassian Rovo Flaw Enabled Data Theft via Prompt Injection
A one-click prompt injection attack on Atlassian's Rovo could expose enterprise data across connected apps.
SecurityCritical Command Injection Hits MSI RadiX Router
A critical command injection vulnerability in the MSI RadiX AXE6600 router allows remote attackers to execute arbitrary code and gain root access.
SecurityCritical Command Injection Hits MSI Radix
A critical vulnerability in the MSI Radix AXE6600 router allows remote attackers to gain root access via the wps.cgi interface.