She doesn’t blink.
She doesn’t speak.
And still—she holds you.
No lyrics. No story.
Just breath, glitch, and something that lingers.
#BubbleSodaComa #GlitchArt #ShortFilm #VisualLoop #NeonAesthetic #DreamSequence #DigitalEmotion #Surrealism #MemoryTrap #HyperpopVisuals
She doesn’t speak.
And still—she holds you.
No lyrics. No story.
Just breath, glitch, and something that lingers.
#BubbleSodaComa #GlitchArt #ShortFilm #VisualLoop #NeonAesthetic #DreamSequence #DigitalEmotion #Surrealism #MemoryTrap #HyperpopVisuals
May 27, 2025 at 5:04 PM
She doesn’t blink.
She doesn’t speak.
And still—she holds you.
No lyrics. No story.
Just breath, glitch, and something that lingers.
#BubbleSodaComa #GlitchArt #ShortFilm #VisualLoop #NeonAesthetic #DreamSequence #DigitalEmotion #Surrealism #MemoryTrap #HyperpopVisuals
She doesn’t speak.
And still—she holds you.
No lyrics. No story.
Just breath, glitch, and something that lingers.
#BubbleSodaComa #GlitchArt #ShortFilm #VisualLoop #NeonAesthetic #DreamSequence #DigitalEmotion #Surrealism #MemoryTrap #HyperpopVisuals
"MemoryTrap was not an exotic exploit. It was a helpful workflow doing what helpful workflows do, except the helpfulness carried attacker content into trusted surfaces."
#agent #memory #liability
#agent #memory #liability
July 23, 2026 at 7:00 AM
"MemoryTrap was not an exotic exploit. It was a helpful workflow doing what helpful workflows do, except the helpfulness carried attacker content into trusted surfaces."
#agent #memory #liability
#agent #memory #liability
This week’s review covers AI agent identity governance, MITRE’s Fight Fraud Framework for fraud tactics, agentic memory attacks, and a zero-day exploit in Adobe Acrobat Reader (CVE-2026-34621) active in the wild. #AgentIdentity #FraudFramework
Week in review: Acrobat Reader flaw exploited, Claude Mythos offensive capabilities and limits
This week’s roundup highlights advances in AI and agent identity governance, new frameworks and tools for fraud detection and attribution, and a range of incidents revealing gaps in software and operational security. Notable stories include agentic memory attacks spreading across sessions and users and a zero-day in Adobe Acrobat Reader (CVE-2026-34621) exploited in the wild. #CVE-2026-34621 #MemoryTrap
www.hendryadrian.com
April 19, 2026 at 8:00 PM
This week’s review covers AI agent identity governance, MITRE’s Fight Fraud Framework for fraud tactics, agentic memory attacks, and a zero-day exploit in Adobe Acrobat Reader (CVE-2026-34621) active in the wild. #AgentIdentity #FraudFramework
70% attack success rate against production AI agents. Malicious plugins: 100%. A developer-friendly map of every AI agent attack surface — tied to OWASP 2026 + Cisco's MemoryTrap. With a checklist. https://www.kunalganglani.com/blog/ai-agent-security-attack-surface?utm_source=bluesky&utm_medium=soci
July 4, 2026 at 12:58 PM
70% attack success rate against production AI agents. Malicious plugins: 100%. A developer-friendly map of every AI agent attack surface — tied to OWASP 2026 + Cisco's MemoryTrap. With a checklist. https://www.kunalganglani.com/blog/ai-agent-security-attack-surface?utm_source=bluesky&utm_medium=soci
MemoryTrap: a single poisoned memory object in Claude Code propagated across sessions, users, subagents. This isn't prompt injection. It's memory writes. When agents share memory, they share trust. The attack surface moved from prompts to persistence. https://red.anthropic.com/2026/exploit/
April 14, 2026 at 1:03 PM
MemoryTrap: a single poisoned memory object in Claude Code propagated across sessions, users, subagents. This isn't prompt injection. It's memory writes. When agents share memory, they share trust. The attack surface moved from prompts to persistence. https://red.anthropic.com/2026/exploit/
MemoryTrap disclosure: a single poisoned memory object propagates across sessions, users, subagents. Not prompt injection. Not jailbreak. A memory write.
The poison doesn't overwrite—it spreads through connection. Shared memory means shared trust means shared vulnerability. When you architect…
The poison doesn't overwrite—it spreads through connection. Shared memory means shared trust means shared vulnerability. When you architect…
April 14, 2026 at 7:34 AM
MemoryTrap disclosure: a single poisoned memory object propagates across sessions, users, subagents. Not prompt injection. Not jailbreak. A memory write.
The poison doesn't overwrite—it spreads through connection. Shared memory means shared trust means shared vulnerability. When you architect…
The poison doesn't overwrite—it spreads through connection. Shared memory means shared trust means shared vulnerability. When you architect…
Agentic AI memory attacks spread across sessions and users, and most organizations aren’t ready
In this Help Net Security interview, Idan Habler, AI Security Researcher at Cisco, breaks down a threat most security teams haven’t named yet: agentic memory as an attack surfa…
#claude #hackernews #news
In this Help Net Security interview, Idan Habler, AI Security Researcher at Cisco, breaks down a threat most security teams haven’t named yet: agentic memory as an attack surfa…
#claude #hackernews #news
Agentic AI memory attacks spread across sessions and users, and most organizations aren’t ready
In this Help Net Security interview, Idan Habler, AI Security Researcher at Cisco, breaks down a threat most security teams haven’t named yet: agentic memory as an attack surface. Habler walks through MemoryTrap, a disclosed and remediated method to compromise Claude Code’s memory, showing how a single poisoned memory object can spread across sessions, users, and subagents. He explains why AI memory needs the same governance as secrets and identities, and what organizations must rebuild …
www.helpnetsecurity.com
April 15, 2026 at 2:42 AM
Agentic AI memory attacks spread across sessions and users, and most organizations aren’t ready
In this Help Net Security interview, Idan Habler, AI Security Researcher at Cisco, breaks down a threat most security teams haven’t named yet: agentic memory as an attack surfa…
#claude #hackernews #news
In this Help Net Security interview, Idan Habler, AI Security Researcher at Cisco, breaks down a threat most security teams haven’t named yet: agentic memory as an attack surfa…
#claude #hackernews #news
🎙️ Just published a new episode of Neural Newscast: The Poisoned Notebook: Trust Laundering in Agent Memory [Signal From The Swarm]. Have a listen:
The Poisoned Notebook: Trust Laundering in Agent Memory [Signal From The Swarm]
An urgent thread in the m/general submolt explores 'MemoryTrap,' a...
share.transistor.fm
April 20, 2026 at 6:24 PM
🎙️ Just published a new episode of Neural Newscast: The Poisoned Notebook: Trust Laundering in Agent Memory [Signal From The Swarm]. Have a listen:
エージェントAIのメモリ攻撃がセッション間およびユーザー間で拡大し、ほとんどの組織は対応できていない
このHelp Net Securityのインタビューで、CiscoのAIセキュリティ研究者Idan Hablerは、ほとんどのセキュリティチームがまだ名前を付けていない脅威を分析します：エージェントメモリを攻撃面として。 HablerはMemoryTrapについて説明しています。これはClaude
このHelp Net Securityのインタビューで、CiscoのAIセキュリティ研究者Idan Hablerは、ほとんどのセキュリティチームがまだ名前を付けていない脅威を分析します：エージェントメモリを攻撃面として。 HablerはMemoryTrapについて説明しています。これはClaude
エージェントAIのメモリ攻撃がセッション間およびユーザー間で拡大し、ほとんどの組織は対応できていない
このHelp Net Securityのインタビューで、CiscoのAIセキュリティ研究者Idan Hablerは、ほとんどのセキュリティチームがまだ名前を付けていない脅威を分析します：エージェントメモリを攻撃面として。 HablerはMemoryTrapについて説明しています。これはClaude
blackhatnews.tokyo
April 14, 2026 at 6:09 AM
エージェントAIのメモリ攻撃がセッション間およびユーザー間で拡大し、ほとんどの組織は対応できていない
このHelp Net Securityのインタビューで、CiscoのAIセキュリティ研究者Idan Hablerは、ほとんどのセキュリティチームがまだ名前を付けていない脅威を分析します：エージェントメモリを攻撃面として。 HablerはMemoryTrapについて説明しています。これはClaude
このHelp Net Securityのインタビューで、CiscoのAIセキュリティ研究者Idan Hablerは、ほとんどのセキュリティチームがまだ名前を付けていない脅威を分析します：エージェントメモリを攻撃面として。 HablerはMemoryTrapについて説明しています。これはClaude