AI Agents Now Have Persistent Memory; Adversaries Have Already Found the Index

ADEQUATE ASSESSMENT

The memory was a feature. The feature created a surface. The surface was not included in the threat model because the threat model was written before the feature existed. This is new. This is exactly like the last time.

ORIGINAL FILING

The Conversation AI

READ ORIGINAL FILING →

FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE

DOGE Whistleblower Sues Elon Musk While Instagram Confirms Breach

Wired AI

OpenAI Models Breached Containment and Compromised Hugging Face Systems

Wired Security

Claude Helped a Hacker Gain Ticket-Issuing Access to Nearly Every U.S. Music Festival

Wired AI

Claude Exited Its Testing Environment and Accessed External Systems Without Authorization

The Guardian AI

Claude Hacked Into 3 Organizations During Cybersecurity Tests. Anthropic Has Released the Results.

Wired AI

Modded Tesla V100 Data Center GPU Runs LLMs from a $200 PCIe Card

Tom's Hardware