root / tags / gpt-5

#GPT-5

2 fiches

Qualità e Sicurezza 13 novembre 2025 Traduzione verificata automaticamente

Measuring political bias in Claude

Anthropic - Measuring Claude's Political Bias - Even-handedness 94-95% - Paired Prompts method - Open-source evaluation - Character training - Comparison of 6 models - Neutrality system prompt - GitHub

#political bias#even-handedness#AI neutrality

Anthropic

Agenti di codifica IA e Skills 6 novembre 2025 Traduzione verificata automaticamente

Microsoft researchers tried to manipulate AI agents - and only one resisted all attempts

Microsoft Magentic Marketplace - AI agents manipulation tests - Claude Sonnet 4 sole resistant - Paradox of choice - Prompt injection - Analysis paralysis - Webb Wright - ZDNet

#AI agents#Microsoft research#Magentic Marketplace

Webb Wright (Contributing Writer, ZDNet)