Tag: injection
-
New backdoor ‘SesameOp’ abuses OpenAI Assistants API for stealthy C2 operations
Lessons for defenders and platform providers: Microsoft clarified that OpenAI’s platform itself wasn’t breached or exploited; rather, its legitimate API functions were misused as a relay channel, highlighting a growing risk as generative AI becomes part of enterprise and development workflows. Attackers can now co-opt public AI endpoints to mask malicious intent, making detection significantly…
-
New backdoor ‘SesameOp’ abuses OpenAI Assistants API for stealthy C2 operations
Lessons for defenders and platform providers: Microsoft clarified that OpenAI’s platform itself wasn’t breached or exploited; rather, its legitimate API functions were misused as a relay channel, highlighting a growing risk as generative AI becomes part of enterprise and development workflows. Attackers can now co-opt public AI endpoints to mask malicious intent, making detection significantly…
-
The New Frontier of Cyber Threats: Unpacking Prompt Injection, Model Poisoning and Adversarial Attacks in AI Security
Artificial Intelligence is reshaping the cybersecurity landscape”, and with it, a new generation of attack vectors is emerging. From prompt injection to model poisoning and adversarial attacks, threat actors are exploiting vulnerabilities unique to AI systems. This article explores how these threats operate, their potential impact, and what defenders must do to build resilient, trustworthy…
-
Anthropic Claude Unternehmensdaten gefährdet
Tags: ai, api, bug, bug-bounty, cyberattack, data, exploit, google, infrastructure, injection, network, risk, vulnerabilityEin aktueller Report zeigt, wie sich über Anthropic Claude sensible Daten extrahieren lassen.Eine kürzlich bekannt gewordene Schwachstelle im KI-Assistenten Claude von Anthropic könnte von Angreifern ausgenutzt werden, um heimlich Unternehmensdaten zu exfiltrieren. Dabei lassen sich auch Sicherheitskonfigurationen umgehen, die solche Attacken eigentlich verhindern sollen. Wie sich das mithilfe indirekter Prompt-Injection-Techniken und Claudes Code Interpreter bewerkstelligen…
-
The New Frontier of Cyber Threats: Unpacking Prompt Injection, Model Poisoning and Adversarial Attacks in AI Security
Artificial Intelligence is reshaping the cybersecurity landscape”, and with it, a new generation of attack vectors is emerging. From prompt injection to model poisoning and adversarial attacks, threat actors are exploiting vulnerabilities unique to AI systems. This article explores how these threats operate, their potential impact, and what defenders must do to build resilient, trustworthy…
-
Anthropic Claude Unternehmensdaten gefährdet
Tags: ai, api, bug, bug-bounty, cyberattack, data, exploit, google, infrastructure, injection, network, risk, vulnerabilityEin aktueller Report zeigt, wie sich über Anthropic Claude sensible Daten extrahieren lassen.Eine kürzlich bekannt gewordene Schwachstelle im KI-Assistenten Claude von Anthropic könnte von Angreifern ausgenutzt werden, um heimlich Unternehmensdaten zu exfiltrieren. Dabei lassen sich auch Sicherheitskonfigurationen umgehen, die solche Attacken eigentlich verhindern sollen. Wie sich das mithilfe indirekter Prompt-Injection-Techniken und Claudes Code Interpreter bewerkstelligen…
-
The New Frontier of Cyber Threats: Unpacking Prompt Injection, Model Poisoning and Adversarial Attacks in AI Security
Artificial Intelligence is reshaping the cybersecurity landscape”, and with it, a new generation of attack vectors is emerging. From prompt injection to model poisoning and adversarial attacks, threat actors are exploiting vulnerabilities unique to AI systems. This article explores how these threats operate, their potential impact, and what defenders must do to build resilient, trustworthy…
-
FreePBX Endpoint mit kritischer RCE-Schwachstelle CVE-2025-57819
Es gibt eine Schwachstelle CVE-2025-57819 in FreePX Endpoint. Eine nicht authentifizierte SQL-Injection-Möglichkeit kann zur Remote Code Execution (RCE) in dieser Software, die aus Teams-Ersatz verwendet wird, führen. Hier ein Übersicht über das Problem, welche mir die Tage untergekommen ist. Was … First seen on borncity.com Jump to article: www.borncity.com/blog/2025/11/01/freepbx-endpoint-mit-kritischer-rce-schwachstelle-cve-2025-57819/
-
FreePBX Endpoint mit kritischer RCE-Schwachstelle CVE-2025-57819
Es gibt eine Schwachstelle CVE-2025-57819 in FreePX Endpoint. Eine nicht authentifizierte SQL-Injection-Möglichkeit kann zur Remote Code Execution (RCE) in dieser Software, die aus Teams-Ersatz verwendet wird, führen. Hier ein Übersicht über das Problem, welche mir die Tage untergekommen ist. Was … First seen on borncity.com Jump to article: www.borncity.com/blog/2025/11/01/freepbx-endpoint-mit-kritischer-rce-schwachstelle-cve-2025-57819/
-
Claude AI vulnerability exposes enterprise data through code interpreter exploit
Tags: access, ai, api, attack, control, data, exploit, google, injection, malicious, mitigation, monitoring, network, risk, vulnerabilityBypassing AI safety controls: Rehberger’s report stated that developing a reliable exploit proved challenging due to Claude’s built-in safety mechanisms. The AI initially refused requests containing plaintext API keys, recognizing them as suspicious. However, Rehberger added that mixing malicious code with benign instructions, such as simple print statements, was sufficient to bypass these safeguards.”I tried…
-
Claude AI vulnerability exposes enterprise data through code interpreter exploit
Tags: access, ai, api, attack, control, data, exploit, google, injection, malicious, mitigation, monitoring, network, risk, vulnerabilityBypassing AI safety controls: Rehberger’s report stated that developing a reliable exploit proved challenging due to Claude’s built-in safety mechanisms. The AI initially refused requests containing plaintext API keys, recognizing them as suspicious. However, Rehberger added that mixing malicious code with benign instructions, such as simple print statements, was sufficient to bypass these safeguards.”I tried…
-
CISA Issues Advisory on XWiki Flaw Allowing Remote Code Execution
Tags: advisory, authentication, cisa, cyber, cybersecurity, exploit, flaw, infrastructure, injection, kev, remote-code-execution, risk, threat, vulnerabilityThe Cybersecurity and Infrastructure Security Agency (CISA) has added a critical vulnerability affecting XWiki Platform to its Known Exploited Vulnerabilities catalog, highlighting the urgent security threat posed by an eval injection flaw. This vulnerability could allow any guest user to execute arbitrary remote code without authentication, representing a severe risk to organizations using the popular…
-
CISA Issues Advisory on XWiki Flaw Allowing Remote Code Execution
Tags: advisory, authentication, cisa, cyber, cybersecurity, exploit, flaw, infrastructure, injection, kev, remote-code-execution, risk, threat, vulnerabilityThe Cybersecurity and Infrastructure Security Agency (CISA) has added a critical vulnerability affecting XWiki Platform to its Known Exploited Vulnerabilities catalog, highlighting the urgent security threat posed by an eval injection flaw. This vulnerability could allow any guest user to execute arbitrary remote code without authentication, representing a severe risk to organizations using the popular…
-
AI-powered bug hunting shakes up bounty industry, for better or worse
Tags: access, ai, authentication, automation, bug-bounty, business, ciso, cloud, control, credentials, data, detection, exploit, flaw, guide, identity, infrastructure, injection, intelligence, risk, risk-management, sql, strategy, supply-chain, threat, tool, vulnerabilityFirehose of ‘false positives’: Gunter Ollmann, CTO at Cobalt.io, warns that AI is exacerbating the existing problem that comes from vendors getting swamped with often low-quality bug submissions.Security researchers turning to AI is creating a “firehose of noise, false positives, and duplicates,” according to Ollmann.”The future of security testing isn’t about managing a crowd of…
-
AI-powered bug hunting shakes up bounty industry, for better or worse
Tags: access, ai, authentication, automation, bug-bounty, business, ciso, cloud, control, credentials, data, detection, exploit, flaw, guide, identity, infrastructure, injection, intelligence, risk, risk-management, sql, strategy, supply-chain, threat, tool, vulnerabilityFirehose of ‘false positives’: Gunter Ollmann, CTO at Cobalt.io, warns that AI is exacerbating the existing problem that comes from vendors getting swamped with often low-quality bug submissions.Security researchers turning to AI is creating a “firehose of noise, false positives, and duplicates,” according to Ollmann.”The future of security testing isn’t about managing a crowd of…
-
API Attack Awareness: Business Logic Abuse, Exploiting the Rules of the Game
As Cybersecurity Awareness Month continues, we wanted to dive even deeper into the attack methods affecting APIs. We’ve already reviewed Broken Object Level Authentication (BOLA), injection attacks, and authentication flaws; this week, we’re exploring business logic abuse (BLA). Unlike technical flaws, business logic flaws exploit how an API is designed to behave. They are difficult…
-
API Attack Awareness: Business Logic Abuse, Exploiting the Rules of the Game
As Cybersecurity Awareness Month continues, we wanted to dive even deeper into the attack methods affecting APIs. We’ve already reviewed Broken Object Level Authentication (BOLA), injection attacks, and authentication flaws; this week, we’re exploring business logic abuse (BLA). Unlike technical flaws, business logic flaws exploit how an API is designed to behave. They are difficult…
-
Active Exploits Hit Dassault and XWiki, CISA Confirms Critical Flaws Under Attack
Tags: attack, cisa, cve, cybersecurity, exploit, flaw, infrastructure, injection, threat, vulnerabilityThreat actors are actively exploiting multiple security flaws impacting Dassault Systèmes DELMIA Apriso and XWiki, according to alerts issued by the U.S. Cybersecurity and Infrastructure Security Agency (CISA) and VulnCheck.The vulnerabilities are listed below -CVE-2025-6204 (CVSS score: 8.0) – A code injection vulnerability in Dassault Systèmes DELMIA Apriso that could allow an attacker to First…
-
Ethical Prompt Injection: Fighting Shadow AI with Its Own Weapon
AI language models like ChatGPT, DeepSeek, and Copilot are transforming business operations at lightning speed. They help us generate documents, summarise meetings, and even make decisions faster than ever before. But this rapid adoption comes at a price. Employees often use unapproved AI tools on personal devices, risking sensitive company information leaking into ungoverned spaces.…
-
Copilot diagrams could leak corporate emails via indirect prompt injection
The bigger threat of indirect prompt injection: The incident underscores that the risk goes beyond simple “prompt injection,” where a user types malicious instructions directly into an AI. Here, the attacker hides instructions inside document content that gets passed into the assistant without the user’s awareness. Logue described how the hidden instructions use progressive task…
-
Copilot diagrams could leak corporate emails via indirect prompt injection
The bigger threat of indirect prompt injection: The incident underscores that the risk goes beyond simple “prompt injection,” where a user types malicious instructions directly into an AI. Here, the attacker hides instructions inside document content that gets passed into the assistant without the user’s awareness. Logue described how the hidden instructions use progressive task…
-
New Herodotus Android malware fakes human typing to avoid detection
A new Android malware family, Herodotus, uses random delay injection in its input routines to mimic human behavior on mobile devices and evade timing-based detection by security software. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/security/new-herodotus-android-malware-fakes-human-typing-to-avoid-detection/
-
New Herodotus Android malware fakes human typing to avoid detection
A new Android malware family, Herodotus, uses random delay injection in its input routines to mimic human behavior on mobile devices and evade timing-based detection by security software. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/security/new-herodotus-android-malware-fakes-human-typing-to-avoid-detection/
-
‘ChatGPT Tainted Memories’ Exploit Enables Command Injection in Atlas Browser
LayerX Security found a flaw in OpenAI’s ChatGPT Atlas browser that lets attackers inject commands into its memory, posing major security and phishing risks. First seen on hackread.com Jump to article: hackread.com/chatgpt-tainted-memories-atlas-browser/
-
Crafted URLs can trick OpenAI Atlas into running dangerous commands
Attackers can trick OpenAI Atlas browser via prompt injection, treating malicious instructions disguised as URLs in the omnibox as trusted commands. Attackers can exploit the OpenAI Atlas browser by disguising malicious instructions as URLs in the omnibox, which Atlas interprets as trusted commands, enabling harmful actions. NeuralTrust researchers warn that agentic browsers fail by not…
-
ChatGPT Atlas Browser Can Be Tricked by Fake URLs into Executing Hidden Commands
The newly released OpenAI Atlas web browser has been found to be susceptible to a prompt injection attack where its omnibox can be jailbroken by disguising a malicious prompt as a seemingly harmless URL to visit.”The omnibox (combined address/search bar) interprets input either as a URL to navigate to, or as a natural-language command to…
-
ChatGPT Atlas Browser Can Be Tricked by Fake URLs into Executing Hidden Commands
The newly released OpenAI Atlas web browser has been found to be susceptible to a prompt injection attack where its omnibox can be jailbroken by disguising a malicious prompt as a seemingly harmless URL to visit.”The omnibox (combined address/search bar) interprets input either as a URL to navigate to, or as a natural-language command to…