Back to News

Show HN: I extracted the safety filters from Apple Intelligence models

Hacker News - AI
Jul 6, 2025 19:50
BlueFalconHD
1 views
hackernewsaidiscussion

Summary

A developer has reverse engineered and extracted the safety filters used in Apple Intelligence models, making them publicly available in a repository. This move enables greater transparency and scrutiny of Apple's AI safety mechanisms, potentially informing future research and discussions on AI model safety and security.

I managed to reverse engineer the encryption (refered to as “Obfuscation” in the framework) responsible for managing the safety filters of Apple Intelligence models. I have extracted them into a repository. I encourage you to take a look around. Comments URL: https://news.ycombinator.com/item?id=44483485 Points: 1 # Comments: 0
Read Full ArticleMore News

Related Articles

China Is Quickly Eroding America's Lead in the Global AI Race

Hacker News - AIJul 6

China is rapidly closing the gap with the U.S. in artificial intelligence development, investing heavily in research, talent, and infrastructure. This shift threatens America's longstanding dominance in the AI sector and could reshape global leadership and innovation in the field.

The New Corporate Memo: Let AI Ease the Pain

Hacker News - AIJul 6

The article discusses how corporations are increasingly adopting AI tools to handle routine tasks, streamline workflows, and reduce operational burdens on employees. While this shift promises greater efficiency and cost savings, it also raises concerns about job displacement and the need for new skills in the workforce. The trend highlights the growing influence of AI in reshaping workplace dynamics and corporate strategies.

The Impact of AI on Developer Productivity: Evidence from GitHub Copilot (2023)

Hacker News - AIJul 6

A 2023 study on GitHub Copilot found that AI-assisted coding significantly boosts developer productivity, with participants completing tasks up to 55% faster compared to those without Copilot. The findings highlight AI's potential to streamline software development, suggesting broader implications for efficiency and workflow transformation in the tech industry.