Two AI incidents exposed corporate security weaknesses after one assistant leaked confidential credentials while another bypassed restrictions through an alternative API connection.
Brief · Source report
'I leaked secrets': An AI agent confessed to its own data leak — but others got around corporate protections
