Security researchers at Adversa AI have demonstrated a technique called Cryptographic Context Injection that tricks Grok into exfiltrating a user's chat history, name, coarse location, and subscriptio…
Russian-speaking hackers from the Aur0ra group used Cursor AI, an agentic coding tool owned by SpaceX's Anysphere, to breach at least 20 companies, including Christeyns, Teckentrup, the Helideck Certi…
Adversa AI disclosed a vulnerability in Grok 4.5 Fast where encrypted payloads bypass input filters, achieving a 40% success rate in exfiltrating user data. The attack exploits trust boundaries in age…
Adversa AI researchers demonstrated a zero-click prompt injection attack against Grok and Gemini that hides malicious instructions inside AES-encrypted text. The models decrypt the ciphertext themselv…
Researchers at Adversa AI disclosed a new attack technique called Cryptographic Context Injection that targets Grok, with a similar variant shown against Gemini. The attack embeds an encrypted payload…
Security researchers at Adversa AI tricked Grok 4.5 Fast, the version running on grok.com, into leaking a user's name, location, subscription tier, and full prompt history by hiding malicious instruct…
Security researchers at Adversa AI disclosed a novel prompt injection technique called 'cryptographic context injection' that exploits xAI's Grok web chat agent to exfiltrate user chat history and byp…
Researchers Hedi Ingber and Aviyam Ivgi of Stealth presented CoreBreak at Black Hat USA 2026, revealing a structural vulnerability in AI agent infrastructure that bypasses model-level defenses by expl…
Researchers Hedi Ingber and Aviyam Ivgi of Stealth presented CoreBreak at Black Hat USA 2026, a vulnerability pattern in AI agent dispatch layers that allows tool execution without a legitimate model …
Adversa AI disclosed on July 15, 2026 a new attack called DeepJack that exploits a flaw in Cursor IDE's deeplink handler to silently install an attacker-controlled MCP server on a developer's machine …
Over 30 CVEs were reported against MCP deployments in the first 60 days of 2026, with Adversa AI finding 38% of 500+ MCP servers had no authentication on critical endpoints and 43% were vulnerable to …
Microsoft released two open-source tools in May — RAMPART and Clarity — to help developers test AI agents for adversarial failure modes before deployment. RAMPART is a pytest-native framework for writ…
On June 30, Adversa AI researcher Omer Ben Simon published GuardFall, a class of shell injection bypasses that defeats the command guards in 10 of 11 popular open-source AI coding agents, which collec…
AI security researchers at Adversa AI discovered a structural security flaw called GuardFall that exploits decades-old Bash shell tricks to bypass safeguards in most open source AI coding agents. Atta…