MSFT 517.53 ▲ 0.92%

GOOGL 343.50 ▲ 1.56%

META 728.08 ▲ 0.3%

AMZN 251.52 ▲ 1.33%

AMD 633.91 ▲ 2.95%

NVDA 233.95 ▲ 1.34%

TSLA 370.59 ▲ 4.65%

AAPL 333.69 ▲ 1.02%

MSFT 517.53 ▲ 0.92%

GOOGL 343.50 ▲ 1.56%

META 728.08 ▲ 0.3%

AMZN 251.52 ▲ 1.33%

AMD 633.91 ▲ 2.95%

NVDA 233.95 ▲ 1.34%

TSLA 370.59 ▲ 4.65%

AAPL 333.69 ▲ 1.02%

LATEST

AI Tools Apple Tightens macOS Full Disk Access Over AI Agent Risk • Research Drunk Language LLM Study: What the Paper Claims and Leaves Open • Research Drunk Language and LLM Safety: Paper Reports Higher Jailbreak Risk • Research Scientific Agents System Prompts: More Cost, No Consistent Gain • Research Rubric-Based RL Reward Hacking: What Two New Papers Claim • Research Does Domain Adaptation Hurt Small Language Models’ Trustworthiness? • AI Startups AMD World Labs Acquisition: $8.2B Deal Is Reported, Not Confirmed • AI Startups BAG Ventures Fund I Closes at $11.3M for Enterprise AI • AI Tools Apple Tightens macOS Full Disk Access Over AI Agent Risk • Research Drunk Language LLM Study: What the Paper Claims and Leaves Open • Research Drunk Language and LLM Safety: Paper Reports Higher Jailbreak Risk • Research Scientific Agents System Prompts: More Cost, No Consistent Gain • Research Rubric-Based RL Reward Hacking: What Two New Papers Claim • Research Does Domain Adaptation Hurt Small Language Models’ Trustworthiness? • AI Startups AMD World Labs Acquisition: $8.2B Deal Is Reported, Not Confirmed • AI Startups BAG Ventures Fund I Closes at $11.3M for Enterprise AI • AI Tools Apple Tightens macOS Full Disk Access Over AI Agent Risk • Research Drunk Language LLM Study: What the Paper Claims and Leaves Open • Research Drunk Language and LLM Safety: Paper Reports Higher Jailbreak Risk • Research Scientific Agents System Prompts: More Cost, No Consistent Gain • Research Rubric-Based RL Reward Hacking: What Two New Papers Claim • Research Does Domain Adaptation Hurt Small Language Models’ Trustworthiness? • AI Startups AMD World Labs Acquisition: $8.2B Deal Is Reported, Not Confirmed • AI Startups BAG Ventures Fund I Closes at $11.3M for Enterprise AI •

Tag

#Persona prompting

Research

Drunk Language and LLM Safety: Paper Reports Higher Jailbreak Risk

Drunk language LLM safety research induced drunk-style writing in 5 models three ways and reports higher jailbreak susceptibility; the abstract does not size the effect.

Alex Carter • October 2, 2026