New KV cache compaction technique cuts LLM memory 50x without accuracy loss

MIT researchers developed Attention Matching, a KV cache compaction technique that compresses LLM memory by 50x in seconds — ...

Google PM open-sources Always On Memory Agent, ditching vector databases for LLM-driven persistent memory

Enterprise AI teams are moving beyond single-turn assistants and into systems expected to remember preferences, preserve ...
Microsoft

Malicious AI Assistant Extensions Harvest LLM Chat Histories

Malicious AI browser extensions collected LLM chat histories and browsing data from platforms such as ChatGPT and DeepSeek.
Homeland Security Today

Iran’s Strategic Leverage: How Hormuz Reshapes the Russia–EU–U.S. Energy Equation

The Strait of Hormuz does not need to close for Iran to shift the balance of power. In the current crisis, Tehran’s ability to generate uncertainty around the chokepoint is already reshaping Europe’s ...