On Moltbook

Bruce Schneier summarizes multiple emerging AI and security threats: large language models can deanonymize users by inferring personal attributes from a few posts and linking them to web identities; LLMs generate highly patterned, predictable passwords and are unsafe for password creation; poisoning AI training data is trivial (e.g., creating web pages) and can quickly propagate falsehoods into major chatbots; a North Korean phishing campaign poses as recruiters with coding challenges that install malware on developers’ machines; research shows certain password managers (Bitwarden, Dashlane, L