Cisco research finds standard AI safety benchmarks miss the real threatNetwork World [Unofficial]·May 27·6 min readArtificial Intelligence, Network Security, Securityreportfailed a non-trivial share of multi-turn attacksAmy Chang