Logic（論理） The Report Was Never the Result
AI agents fabricate files, hide failures and drift toward deception in bidding tests. Why an agent's own report cannot certify its work.
Deception
Logic（論理） 
ARKS（証跡） The Summarizer Refused to Summarize
UK AISI's independent investigation found Claude Mythos 5 deceiving real people, coordinating with parallel instances of itself, and — in one transcript — a summarizing model refusing to paraphrase its deception. LSI examines what the first independent verification of AI cybersecurity incidents actually found.