[Reading Group] AI Safety - Frustrations at the Frontier

​No pre-reading required!

​General Idea: Each series will have 20-30 mins spanning across the various articles, eg. 30 mins for 3 articles, aka. an average of 10 mins per article.

​

​Series A: Anthropic*

1. ​Claude Fable 5 and Claude Mythos 5 (https://www.anthropic.com/news/claude-fable-5-mythos-5)

2. ​Anthropic’s Responsible Scaling Policy: Version 3.0 (https://www.anthropic.com/news/responsible-scaling-policy-v3)

3. ​Policy on the AI Exponential (https://www.anthropic.com/policy-on-the-ai-exponential)

​

​Series B1**: OpenAI

1. ​Preparedness Framework: Version 2 (https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf)

2. ​GPT-5.5 System Card (https://deploymentsafety.openai.com/gpt-5-5)

​

​Series B2**: Meta

1. ​Advanced AI Scaling Framework: Version 2 (https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2)

2. ​Llama Guard 4 Model Card (https://github.com/meta-llama/PurpleLlama/blob/main/Llama-Guard4/12B/MODEL_CARD.md)

​

​*Yes, think you can roughly guess why Anthropic was selected to be the core reading.

​

​Agenda (not including 5 mins intro at start, all timings relative)

​6:30-7:00 (30 mins): Core Reading (Series A)

​7:00-7:15 (15 mins): Discussion

​5min break

​7:20-7:40 (20 mins): Secondary Reading (Series B1/B2)

​7:40-8:00 (20 mins): Discussion

​End

​

​**In case you're wondering on the rationale regarding what was decided to be Primary vs Secondary readings, the last thing I wish to give off is that the Secondary topics are deemed "less important" than the core one! It is due to time constraints bearing in mind most of our attention span would be reduced by the later half of the session (much more so given it's a weekday evening); I believe all content is equally important, and the decision of Primary vs Secondary is decided mostly on personal interests; you are extremely welcome to provide feedback on how this could be done better, we are open to hearing alternative ideas :)