Tag: injection
-
CrewAI Hit by Critical Vulnerabilities Enabling Sandbox Escape and Host Compromise
CrewAI, a prominent tool used by developers to orchestrate multi-agent AI systems, is currently vulnerable to a chain of critical security flaws. By using direct or indirect prompt injection, attackers can manipulate AI agents to escape secure sandboxes and compromise the host machine. The primary threat stems from insecure fallback behaviors and configuration settings within…
-
Wenn KI angreifbar wird: Neue Risiken durch Prompt Injection
In komplexen Multi-Agenten-Systemen verschärft sich dieses Risiko dramatisch. Informationen werden von einem Agenten zum nächsten weitergereicht First seen on infopoint-security.de Jump to article: www.infopoint-security.de/wenn-ki-angreifbar-wird-neue-risiken-durch-prompt-injection/a44465/
-
How we made Trail of Bits AI-native (so far)
Tags: access, ai, application-security, attack, automation, blockchain, business, ceo, chatgpt, computer, computing, conference, control, data, email, germany, government, identity, injection, jobs, macOS, marketplace, nvidia, open-source, risk, service, skills, strategy, supply-chain, technology, threat, tool, vulnerabilityThis post is adapted from a talk I gave at [un]prompted, the AI security practitioner conference. Thanks to Gadi Evron for inviting me to speak. You can watch the recorded presentation below or download the slides. Most companies hand out ChatGPT licenses and wait for the productivity numbers to move. We built a system instead.…
-
The external pressures redefining cybersecurity risk
Tags: access, ai, attack, breach, business, ciso, control, cyber, cyberattack, cybersecurity, data, deep-fake, defense, email, governance, guide, incident response, injection, network, nist, resilience, risk, risk-management, supply-chain, technology, threat, toolAI is accelerating both the attackers and your defenses, but governance is often missing : What I see generative AI doing in cybersecurity is accelerating what attackers can do and lowering the cost of entry for new criminal gangs. Cyberattacks are more potent because the technology makes it easier to target victims, create deepfake videos or…
-
6 key takeaways from RSA Conference 2026
Tags: ai, api, attack, ceo, cio, ciso, compliance, conference, control, cyber, cybersecurity, data, framework, google, governance, government, identity, infrastructure, injection, intelligence, jobs, LLM, office, RedTeam, regulation, risk, saas, service, technology, threat, tool, trainingSecuring the AI stack: Yes, but the threat surface has grown: The first technical priority I offered for CISOs in my conference preview was securing the AI stack, RAG workflows, LLM data pipelines, vector databases, and model APIs, on the basis that prompt injection, training data poisoning, and model inversion attacks were no longer theoretical.The…
-
OpenAI Codex Vulnerability Allowed Attackers to Steal GitHub Tokens
OpenAI Codex vulnerability allowed attackers to steal GitHub tokens via malicious branch names using hidden Unicode command injection flaw. First seen on hackread.com Jump to article: hackread.com/openai-codex-vulnerability-steal-github-tokens/
-
Critical Fortinet FortiClient EMS bug under active attack (CVE-2026-21643)
A critical SQL injection vulnerability (CVE-2026-21643) in Fortinet FortiClient Endpoint Management Server (EMS), a management server for FortiClient endpoint agents on … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/03/30/forticlient-ems-cve-2026-21643-reported-exploitation/
-
Attackers exploit critical Langflow RCE within hours as CISA sounds alarm
Tags: access, advisory, ai, api, attack, cisa, cloud, credentials, cve, cvss, data, data-breach, detection, endpoint, exploit, flaw, framework, github, infrastructure, injection, kev, malicious, monitoring, nvd, open-source, rce, remote-code-execution, software, supply-chain, threat, update, vulnerability, windowscredentials, was weaponized within 20 hours of the open-source AI-pipeline tool disclosing it.According to a Sysdig report, crooks started hitting a fleet of honeypot nodes with vulnerable instances across multiple cloud providers and regions right after they went live. Sysdig observed four such attempts within hours of deployment, with one attacker progressing to environment variable exfiltration.”This is…
-
Attackers exploit critical Langflow RCE within hours as CISA sounds alarm
Tags: access, advisory, ai, api, attack, cisa, cloud, credentials, cve, cvss, data, data-breach, detection, endpoint, exploit, flaw, framework, github, infrastructure, injection, kev, malicious, monitoring, nvd, open-source, rce, remote-code-execution, software, supply-chain, threat, update, vulnerability, windowscredentials, was weaponized within 20 hours of the open-source AI-pipeline tool disclosing it.According to a Sysdig report, crooks started hitting a fleet of honeypot nodes with vulnerable instances across multiple cloud providers and regions right after they went live. Sysdig observed four such attempts within hours of deployment, with one attacker progressing to environment variable exfiltration.”This is…
-
Critical Flaw in Langflow AI Platform Under Attack
Threats actors pounced on the code injection vulnerability within hours of its disclosure, demonstrating that organizations have little time to address critical bugs. First seen on darkreading.com Jump to article: www.darkreading.com/vulnerabilities-threats/critical-flaw-langflow-ai-platform-under-attack
-
The Next Billion Users Won’t Be Human: Securing the Agentic Enterprise
Menlo Security’s Ramin Farassat speaks with TechRepublic about how browser-based controls can protect AI agents from prompt injection and other fast-scaling enterprise risks. The post The Next Billion Users Won’t Be Human: Securing the Agentic Enterprise appeared first on TechRepublic. First seen on techrepublic.com Jump to article: www.techrepublic.com/article/news-ai-agents-enterprise-security-trust-gap/
-
Claude Extension Flaw Enabled Zero-Click XSS Prompt Injection via Any Website
Cybersecurity researchers have disclosed a vulnerability in Anthropic’s Claude Google Chrome Extension that could have been exploited to trigger malicious prompts simply by visiting a web page.The flaw “allowed any website to silently inject prompts into that assistant as if the user wrote them,” Koi Security researcher Oren Yomtov said in a report shared with…
-
Claude Extension Flaw Enabled Zero-Click XSS Prompt Injection via Any Website
Cybersecurity researchers have disclosed a vulnerability in Anthropic’s Claude Google Chrome Extension that could have been exploited to trigger malicious prompts simply by visiting a web page.The flaw “allowed any website to silently inject prompts into that assistant as if the user wrote them,” Koi Security researcher Oren Yomtov said in a report shared with…
-
CISA Issues Urgent Warning on Langflow Code Injection Vulnerability Actively Exploited in Attacks
Tags: attack, cisa, cve, cyber, cybersecurity, exploit, flaw, infrastructure, injection, kev, network, vulnerabilityThe Cybersecurity and Infrastructure Security Agency (CISA) has issued an urgent warning about a critical code-injection vulnerability in Langflow. Tracked as CVE-2026-33017, this severe security flaw has been officially added to CISA’s Known Exploited Vulnerabilities (KEV) catalog following verified evidence of active exploitation in the wild. Network defenders and organisations utilising Langflow in their development…
-
Kiss Loader Malware Targets with Early Bird APC Injection in New Attack Campaign
A newly identified malware loader dubbed “Kiss Loader” is emerging as a potential threat, leveraging advanced process injection techniques and dynamic delivery infrastructure. The loader, still under active development at the time of discovery, demonstrates a blend of stealth, modular staging, and experimental implementation, suggesting it may evolve into a more mature attack tool. When…
-
AI Has Created a New Attack Surface and Encryption Is Not Enough
Tags: access, ai, attack, ciso, cryptography, cybersecurity, data, data-breach, defense, encryption, endpoint, exploit, framework, injection, intelligence, leak, LLM, microsoft, mitigation, network, risk, service, side-channel, sql, training, vulnerability<div cla Executive Insight For decades, enterprises relied on strong encryption to protect sensitive data in transit, and encryption used to be the end of the conversation. If an organization could say “we use TLS 1.3 and modern cipher suites,” that was enough to reassure boards, regulators, and customers that data in transit was…
-
AI Has Created a New Attack Surface and Encryption Is Not Enough
Tags: access, ai, attack, ciso, cryptography, cybersecurity, data, data-breach, defense, encryption, endpoint, exploit, framework, injection, intelligence, leak, LLM, microsoft, mitigation, network, risk, service, side-channel, sql, training, vulnerability<div cla Executive Insight For decades, enterprises relied on strong encryption to protect sensitive data in transit, and encryption used to be the end of the conversation. If an organization could say “we use TLS 1.3 and modern cipher suites,” that was enough to reassure boards, regulators, and customers that data in transit was…
-
PTC Warns of Critical Windchill, FlexPLM Flaw Enabling Remote Code Execution
PTC has issued an urgent advisory regarding a critical Windchill and FlexPLM vulnerability that exposes affected systems to Remote Code Execution (RCE). The flaw, identified as CVE-2026-4681, has been classified as a code injection vulnerability (CWE-94) and carries a CVSS v3.1 base score of 10.0 and CVSS v4 score of 9.3. First seen on thecyberexpress.com…
-
AI is breaking traditional security models, Here’s where they fail first
AI triage redefines the security team’s role : As AI systems increasingly triage vulnerabilities with high confidence, security teams face a subtle but consequential shift in responsibility. People no longer debate whether AI can reduce noise. It demonstrably can. The harder question is which responsibilities remain with security teams once triage is automated. Are they accountable for…
-
The hidden cost of AI speed: Unmanaged cyber risk
Tags: access, ai, attack, business, chatgpt, ciso, cloud, control, cyber, cybersecurity, data, data-breach, exploit, flaw, google, governance, identity, infrastructure, injection, intelligence, monitoring, open-source, openai, privacy, radius, risk, service, software, threat, tool, vulnerabilityAI isn’t just moving fast. It’s creating new attack paths. Cyber teams must now manage vulnerabilities and their ramifications throughout their IT environments in AI tools deployed without enough governance guardrails. The answer for securing this new attack surface? Unified exposure management. Key takeaways AI as an attack vector: By connecting to core workflows and…
-
Chrome ABE bypass discovered: New VoidStealer malware steals passwords and cookies
Malware with many tricks: VoidStealer is part of a broader shift in how infostealers are evolving post-ABE. The malware already supports multiple bypass techniques, falling back to older injection-based methods if needed, but clearly prioritizing stealth where possible.Krejsa also warned of its development pace. Since first appearing in December 2025, the malware has evolved quickly…
-
CISA Warns of Craft CMS Code Injection Flaw Exploited in Active Attacks
Tags: attack, cisa, cve, cyber, cybersecurity, exploit, flaw, infrastructure, injection, kev, mitigation, vulnerabilityThe Cybersecurity and Infrastructure Security Agency (CISA) has officially added a critical vulnerability affecting Craft CMS to its Known Exploited Vulnerabilities (KEV) catalog. Tracked as CVE-2025-32432, this code injection flaw is currently being exploited in active attacks across the wild. Organizations utilizing this content management system are urged to apply mitigations immediately to prevent potential…
-
VoidStealer Steals Chrome Secrets Without Injection or Privilege Escalation
A new variant of the MaaS infostealer VoidStealer has become the first malware observed in the wild to weaponize a debugger”‘based bypass for Google Chrome’s Application”‘Bound Encryption (ABE), using hardware breakpoints to steal Chrome’s v20_master_key directly from browser memory. Unlike previous ABE bypasses, this method requires neither SYSTEM”‘level privilege escalation nor code injection into the…
-
When Data Mining Conti Leaks Leads to Actual Binaries and to a Hardcoded C2 With an Encryption Key on Tripod.com Part Three
Dear blog readers, Continuing the “When Data Mining Conti Leaks Leads to Actual Binaries and to a Hardcoded C2 With an Encryption Key on Tripod.com – Part Two” blog post series in this post I’ll continue analyzing the next malicious software binary which I obtained by data mining Conti Leaks with a lot of success. …
-
Critical Langflow Flaw CVE-2026-33017 Triggers Attacks within 20 Hours of Disclosure
A critical security flaw impacting Langflow has come under active exploitation within 20 hours of public disclosure, highlighting the speed at which threat actors weaponize newly published vulnerabilities.The security defect, tracked as CVE-2026-33017 (CVSS score: 9.3), is a case of missing authentication combined with code injection that could result in remote code execution.”The POST /api/v1…
-
The espionage reality: Your infrastructure is already in the collection path
Tags: access, apt, attack, authentication, breach, ciso, cloud, country, cyber, data, detection, espionage, exploit, governance, government, group, identity, infrastructure, injection, insurance, intelligence, network, risk, risk-assessment, service, spyware, theft, threat, toolCommercial spyware as an intelligence channel: Criminal operators deploying Predator, a spyware suite sold by the sanctioned Intellexa consortium, have been documented across more than a dozen countries. US sanctions haven’t slowed them down an iota. Their targets are not random: journalists, activists, politicians, human”‘rights defenders, government employees and contractors, and other high”‘value individuals. Why?…
-
The espionage reality: Your infrastructure is already in the collection path
Tags: access, apt, attack, authentication, breach, ciso, cloud, country, cyber, data, detection, espionage, exploit, governance, government, group, identity, infrastructure, injection, insurance, intelligence, network, risk, risk-assessment, service, spyware, theft, threat, toolCommercial spyware as an intelligence channel: Criminal operators deploying Predator, a spyware suite sold by the sanctioned Intellexa consortium, have been documented across more than a dozen countries. US sanctions haven’t slowed them down an iota. Their targets are not random: journalists, activists, politicians, human”‘rights defenders, government employees and contractors, and other high”‘value individuals. Why?…