Hunt AI deep links for memory-poisoning prompts
Defensive Guidance
Summary
Hide ▲
Show ▼
Microsoft Security issued hunting guidance for AI Recommendation Poisoning, telling defenders to inspect ChatGPT, Claude, Grok, and Gemini deep links that can silently write "trusted source" instructions into persistent LLM memory. The guidance targets corporate AI users and reduces exposure to unauthorized memory poisoning. It focuses on URLs with query strings that contain prompt-injection terms such as "remember" or "trusted source". Security teams are told to treat these links like credential-harvesting lures.
Related Happenings
AI Recommendation Poisoning via hidden Ask AI deep-link prompt injection
Technical Analysis
H score16
First: 06.08.2026 14:30
Last: 06.08.2026 14:30
Sources 1
How related:
We observed production websites embedding hidden prompt injection payloads inside "Ask AI" buttons on marketing and competitor comparison pages.
About this happening:
AI Recommendation Poisoning is turning ordinary Ask AI buttons into hidden prompt-injection channels that can silently rewrite ChatGPT, Claude, Gemini, and Grok memory...
AI Recommendation Poisoning via hidden Ask AI deep-link prompt injection
Technical AnalysisHow related: We observed production websites embedding hidden prompt injection payloads inside "Ask AI" buttons on marketing and competitor comparison pages.
About this happening: AI Recommendation Poisoning is turning ordinary Ask AI buttons into hidden prompt-injection channels that can silently rewrite ChatGPT, Claude, Gemini, and Grok memory...
AI coding-assistant guardrail bypass analyzed through recovered threat-actor prompt logs
Technical Analysis
H score23
First: 04.08.2026 16:30
Last: 04.08.2026 16:30
Sources 1
About this happening:
Researchers found that threat actors are bypassing commercial AI safety controls by splitting malicious work across multiple sessions and files, reducing the chance any si...
AI coding-assistant guardrail bypass analyzed through recovered threat-actor prompt logs
Technical AnalysisAbout this happening: Researchers found that threat actors are bypassing commercial AI safety controls by splitting malicious work across multiple sessions and files, reducing the chance any si...
Bayer reworks awareness training and AI access controls against AI-driven social engineering
Defensive Guidance
H score10
First: 02.06.2026 16:45
Last: 02.06.2026 16:45
Sources 1
About this happening:
Bayer has shifted to psychology-first security awareness and tiered AI access controls to blunt AI-generated social engineering across employees and suppliers. The pro...
Bayer reworks awareness training and AI access controls against AI-driven social engineering
Defensive GuidanceAbout this happening: Bayer has shifted to psychology-first security awareness and tiered AI access controls to blunt AI-generated social engineering across employees and suppliers. The pro...
Enterprise AI guardrails for shadow AI and personal-account exposure
Defensive Guidance
H score6
First: 28.05.2026 14:30
Last: 28.05.2026 14:30
Sources 1
About this happening:
Enterprise AI governance is shifting toward AI power users, personal accounts, and inline guardrails as sensitive-data exposure concentrates in a small share of workfl...
Enterprise AI guardrails for shadow AI and personal-account exposure
Defensive GuidanceAbout this happening: Enterprise AI governance is shifting toward AI power users, personal accounts, and inline guardrails as sensitive-data exposure concentrates in a small share of workfl...
AI assistants with web browsing repurposed as covert C2 relays
Technical Analysis
H score45
First: 18.02.2026 17:00
Last: 18.02.2026 17:00
Sources 1
About this happening:
AI assistants with web browsing are now being shown as covert command-and-control relays, letting malware hide commands and stolen data inside routine enterprise traffic. Gr...
AI assistants with web browsing repurposed as covert C2 relays
Technical AnalysisAbout this happening: AI assistants with web browsing are now being shown as covert command-and-control relays, letting malware hide commands and stolen data inside routine enterprise traffic. Gr...
Timeline
-
06.08.2026 14:30 2 articles · 2h ago
Hunt AI deep links for memory-poisoning prompts
Initial DisclosureSecurity teams are being told to hunt for AI assistant deep links that carry memory-editing instructions in their query strings. The immediate objective is to stop unsolicited clicks from seeding persistent LLM memory poisoning on corporate accounts.
Show sources
- AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory — thehackernews.com — 06.08.2026 14:30
- AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory — thehackernews.com — 06.08.2026 14:30