AI Safety & Alignment

Today's Stories

1. 1

   OpenAI proposes international institute to protect youth from AI dangers

   OpenAI announced plans for global cooperation on youth AI safety, calling for an international institute to establish safeguards and standards. The ChatGPT maker published its policy positions on AI regulation and political advocacy on June 1-2, emphasizing the need for thoughtful oversight.

   Parents and educators may soon have clearer guidelines and protections when children use AI tools like ChatGPT in schools or at home.

2. 2

   Study reveals widespread bias in AI hiring tools affecting 25,500 job applications

   Researchers found that AI resume screening systems showed bias in 45% of cases when evaluating identical work histories with only minor demographic changes. The study tested 10 different AI models and found that some penalized candidates for attending prestigious universities or having certain backgrounds, inventing professional-sounding excuses to justify lower scores.

   Job seekers using AI-powered application systems may face unfair treatment based on their background rather than qualifications, potentially affecting hiring decisions at many companies.

3. 3

   AI safety experts admit uncertainty about which actions reduce AI risks

   Holden Karnofsky, a veteran AI safety researcher, believes there's a 49% chance his work is making AI dangers worse rather than better. Jesse Clifton stepped down as executive director of the Center on Long-Term Risk in 2025 for similar reasons, highlighting how even experts struggle to know which safety measures are effective.

   The uncertainty among top AI safety researchers suggests that current efforts to make AI systems safer may not be as reliable as the public assumes.

4. 4

   RSM consulting firm earns top Microsoft AI partnership recognition

   RSM US LLP received Microsoft's Frontier Partner badge on June 3, placing it among elite firms helping businesses implement AI at scale. The recognition highlights RSM's alignment with Microsoft's AI strategy and its 'AI-first, human-led' approach to consulting services.

   Medium-sized businesses working with RSM may get access to more advanced AI implementation services and Microsoft AI tools.

5. 5

   Chinese AI model Minimax M3 found to lack typical political censorship

   Researchers testing AI bias found that Minimax's M3 model appears to have no political censorship, unlike other Chinese AI systems that typically restrict discussions about sensitive topics. This discovery came during development of a benchmark to measure Chinese AI political bias.

   Users of Chinese AI models may have different experiences with political discussions depending on which specific AI system they use.