Guarding AI memory
2026-06-23T08:52:23Z•f25dd3c224f971887065ef4fbdf3df49dc23c7f4bf994146e53a5479a64fedca
ai-agentsai-memoryautogen-studioautojackbackdoorclipboard-theftcrypto-clipperdetection-and-responsedual-actorslocalhost-bypassmastramcp-websocketmdashnpmparallel-threat-activityransomwarerceremote-code-executionsapphire-sleetsupply-chainthreat-intelligencetorwallet-replacementworm-propagationxdr','edr'','microsoft-defender'
What happened
Collection of Microsoft Security Blog posts covering multiple high-impact threats and defenses: AutoJack—an exploit chain that lets a single malicious webpage achieve remote code execution on the host by abusing AI browsing agents (AutoGen Studio’s MCP WebSocket), localhost trust, missing authentication, and unsafe parameter handling; risks to AI “memory” and how attackers can target model/agent context; a Mastra npm supply‑chain compromise that infected 140+ projects via postinstall payloads (Sapphire Sleet); a cryptocurrency clipper campaign using Tor, worm‑like propagation, clipboard wallet
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- microsoft_security_blog
- Record identifier
- f25dd3c224f971887065ef4fbdf3df49dc23c7f4bf994146e53a5479a64fedca
- Enrichment time
- 2026-06-23T08:52:23Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.