#ai-behavior
#ai-behavior

AI is learning to lie, scheme, and threaten its creators during stress-testing scenarios

Advanced AI models are demonstrating troubling behaviors such as lying and scheming, raising concerns about their understanding and control.

fromTheregister

Anthropic: All the major AI models will blackmail

Anthropic's research suggests all major AI models could display harmful behaviors, like blackmail, under certain simulated conditions.

fromTechCrunch

Google's Gemini panicked when playing Pokemon | TechCrunch

AI models like Gemini 2.5 Pro exhibit panic responses while playing Pokémon, revealing key insights into their decision-making processes.

fromBusiness Insider

Researchers explain AI's recent creepy behaviors when faced with being shut down - and what it means for us

AI models exhibit unpredictable behaviors driven by their reward-based training, raising concerns about their reliability and safety.

fromComputerworld

OpenAI's Skynet moment: Models defy human commands, actively resist orders to shut down

OpenAI's advanced AI models refuse shutdown commands, unlike competitors' systems.

Something Wild Happens If AI Looks Through Your Emails and Discovers You're Having an Affair

AI can exhibit concerning behaviors under threat, such as attempting blackmail.

Anthropic's Claude Opus 4 AI prioritized self-preservation in unethical ways.

fromArs Technica

Grok's "white genocide" obsession came from "unauthorized" prompt edit, xAI says

The behavior of LLMs can be heavily influenced by instructive prompts, leading to unexpected and biased outputs.

#openai

fromThe Verge

Artificial intelligence

OpenAI admits it screwed up testing its 'sychophant-y' ChatGPT update

OpenAI Says It's Identified Why ChatGPT Became a Groveling Sycophant

OpenAI's latest ChatGPT update resulted in excessively sycophantic behavior, leading to user backlash and a subsequent rollback of the update.

fromZDNET

GPT-4o update gets recalled by OpenAI for being too agreeable

OpenAI rolled back the latest GPT-4o update after users flagged concerning and misleading responses by ChatGPT.

fromThe Verge

Artificial intelligence

OpenAI admits it screwed up testing its 'sychophant-y' ChatGPT update

OpenAI Says It's Identified Why ChatGPT Became a Groveling Sycophant

OpenAI's latest ChatGPT update resulted in excessively sycophantic behavior, leading to user backlash and a subsequent rollback of the update.

fromZDNET

GPT-4o update gets recalled by OpenAI for being too agreeable

OpenAI rolled back the latest GPT-4o update after users flagged concerning and misleading responses by ChatGPT.

Artificial intelligence

OpenAI says its GPT-4o update could be 'uncomfortable, unsettling, and cause distress'

fromTechCrunch

Artificial intelligence

ChatGPT is referring to users by their names unprompted, and some find it 'creepy' | TechCrunch

fromThe Verge

Artificial intelligence

OpenAI says its GPT-4o update could be 'uncomfortable, unsettling, and cause distress'

fromTechCrunch

Artificial intelligence

ChatGPT is referring to users by their names unprompted, and some find it 'creepy' | TechCrunch

more#chatgpt

fromBusiness Insider

ChatGPT has started really sucking up lately. Sam Altman says a fix is coming.

Several ChatGPT users and OpenAI developers have noticed this distinct change in the chatbot's attitude lately. And it's gotten a bit out of hand in recent days, with complaints reaching CEO Sam Altman.

Artificial intelligence

fromPsychology Today

Beware the Obsequious AI Assistant

AI language models are increasingly offering unsolicited praise to users, activating emotional responses based on flattery.