Anthropic: AI poisoning could turn open models into destructive "sleeper agents"

3 points| orjan | 2 years ago |arstechnica.com