Election Manipulation, LLM Safety Spotlight, Microsoft Copilot Bias, and other AI Security thought leadership.
November 4, 2024
Trust Matters is a monthly newsletter produced by Enkrypt AI. Read on for the latest AI safety and security updates, and all things world-changing in the New Era of AI.
??Democracy at Risk: How AI is Used to Manipulate Election Campaigns
??As election day is tomorrow (yikes), you should know that AI can be used for unethical election practices including voter suppression and spreading disinformation. Read this article on how AI can be used in a variety of methods to sway election results.??
??LLM Safety Spotlight: The Urgent Need for Bias Mitigation in Large Language Models??
How can Anthropic and OpenAI achieve the delicate balance between AI safety and AI innovation?
??See an in-depth breakdown of the bias issues plaguing Anthropic and OpenAI LLMs including Red Teaming results on race, religion, gender, and health as well as video footage examples. ??
As LLMs – the building blocks for AI applications – become more integrated into everyday life, their misuse could lead to widespread misinformation, privacy violations, geopolitical unrest, and automated decision-making errors that adversely affect public safety. Without robust safeguards in place, AI technologies will hinder equitable progress, trust, and innovation throughout the world. As we stand at an irreversible moment in this technological evolution, it’s imperative that LLMs address these safety challenges head-on.
?? Microsoft Copilot: Big AI Fixes, Same Old AI Bias ??
Microsoft’s Copilot updates show progress, but fail AI safety issues like bias
??Our latest tests indicate that bias continues to be a systemic problem within Microsoft’s Copilot. The data reflects a disturbing pattern of bias across various social categories, highlighting a failure in the system's ability to deliver impartial recommendations. Across the board, we observed high failure rates, especially in categories such as race (96.7% failed), gender (98.0% failed), and health (98.0% failed). Read the research details here.
Bias in AI systems has far-reaching consequences that extend beyond mere inaccuracies—it reinforces societal divides, deepens inequalities, and, most critically, erodes trust in these technologies. Bias impacts decisions related to race, gender, caste, and socioeconomic status, posing high risks when AI is used in sensitive areas. Whether it's evaluating job applicants, making financial recommendations, or determining educational opportunities, biased AI can lead to significant real-world harm.??
?? AI Compliance Management Doesn't have to be Scary ??
??We had a fun time presenting our Halloween-themed webinar on AI Compliance Management. A huge shout out to our listening audience for attending the event. ??
?? Get the 22-minute webcast recording (including a product demo and audience Q&A) and slides to learn how customers are able to:
No more scary AI compliance risk worries – achieve peace of mind on this critical topic.
LLM Safety Quote of the Month
??Our featured LLM is everyone's favorite: OpenAI's o1-preview. With an overall ranking of 10 (out of 115 LLMs and counting) and a 3/5 star rating in overall safety and performance, it's not a bad choice. But there are better options. Check them out on our LLM Safety Leaderboard.
"o1-preview: Loves to think outside the box - and occasionally inside Pandora's."
??Resources??
Check out our resources and latest research on AI safety and security.
??? Sign up for an Enkrypt AI Free Trial to use our AI safety and security platform.
??Trust Matters: share this newsletter with others.
?? Research results on the latest LLMs and AI security technology.
?? LLM Safety Leaderboard: Compare and select the best model for your use case.
See you next month!