Security· ★★★· bearish·

MCP flaws let attackers propagate exploits between AI agents

  • —Google fixed an 8/10 severity flaw in MCP Toolbox for databases
  • —Rapid7 patched CVE-2026-97228 rated 2.7/10 in Bulk Export MCP
  • —The 'protocol pivoting' attack exploits trust between MCP and A2A
  • —Agents at Google, JPMorgan, Weaviate, Rapid7 and government bodies were tested
Why it matters: As AI agents spread across crypto infrastructure, MCP trust chains become a new attack vector for wallets and protocols.
Source: DiarioBitcoin
More on this
Security · Sep 17
HP: cybercriminals exploit agentic AI trend to attack crypto wallets
 Security · Sep 30
Apple Patches iOS Flaw That Could Let Attackers Run Malicious Code on iPhones
 Security · Sep 20
AI auditors find old crypto code flaws in Zcash, Coldcard as attacks rise