Week 2: EU Launches General-Purpose AI Code of Practice
Jul 17, 2025
From new regulations to real-world applications, AI continues to shape industries and policy at a rapid pace. This week's signal explores five key developments: the EU’s new voluntary Code of Practice for general-purpose AI, the growing accessibility of running LLMs locally, evolving design patterns for building AI agents, a $5B investment by AWS and HUMAIN to boost AI adoption in Saudi Arabia, and Netflix’s first use of generative AI for visual effects, highlighting both the promise and tensions emerging as AI becomes more embedded in daily life.
European Commission launches the general-purpose AI code of practice
AI Summary
The General-Purpose AI (GPAI) Code of Practice, published on July 10, 2025, is a voluntary tool for industry to comply with the AI Act’s obligations for general-purpose AI models. The Code, consisting of chapters on Transparency, Copyright, and Safety and Security, aims to reduce administrative burden and provide legal certainty for AI model providers.
European Commission launches the general-purpose AI code of practice
AI Summary
The General-Purpose AI (GPAI) Code of Practice, published on July 10, 2025, is a voluntary tool for industry to comply with the AI Act’s obligations for general-purpose AI models. The Code, consisting of chapters on Transparency, Copyright, and Safety and Security, aims to reduce administrative burden and provide legal certainty for AI model providers.
European Commission launches the general-purpose AI code of practice
AI Summary
The General-Purpose AI (GPAI) Code of Practice, published on July 10, 2025, is a voluntary tool for industry to comply with the AI Act’s obligations for general-purpose AI models. The Code, consisting of chapters on Transparency, Copyright, and Safety and Security, aims to reduce administrative burden and provide legal certainty for AI model providers.
MIT Technology Review: How to run an LLM on your laptop
AI Summary
Running local LLMs on personal devices is becoming increasingly accessible, offering privacy, control, and a deeper understanding of AI limitations. While online LLMs are more powerful, local models provide a consistent and customizable experience. Tools like Ollama and LM Studio simplify the process, making local LLMs accessible to users without coding expertise.
MIT Technology Review: How to run an LLM on your laptop
AI Summary
Running local LLMs on personal devices is becoming increasingly accessible, offering privacy, control, and a deeper understanding of AI limitations. While online LLMs are more powerful, local models provide a consistent and customizable experience. Tools like Ollama and LM Studio simplify the process, making local LLMs accessible to users without coding expertise.
MIT Technology Review: How to run an LLM on your laptop
AI Summary
Running local LLMs on personal devices is becoming increasingly accessible, offering privacy, control, and a deeper understanding of AI limitations. While online LLMs are more powerful, local models provide a consistent and customizable experience. Tools like Ollama and LM Studio simplify the process, making local LLMs accessible to users without coding expertise.
Anthropic: Building effective agents
AI Summary
This piece describes two patterns for building LLM applications: evaluator-optimizer and agents. The evaluator-optimizer pattern involves an evaluator LLM providing feedback to an optimizer LLM, enabling iterative refinement. The agent pattern involves an LLM using tools to accomplish tasks autonomously, with human oversight at checkpoints. Both patterns are customizable and can be combined to fit different use cases.
Anthropic: Building effective agents
AI Summary
This piece describes two patterns for building LLM applications: evaluator-optimizer and agents. The evaluator-optimizer pattern involves an evaluator LLM providing feedback to an optimizer LLM, enabling iterative refinement. The agent pattern involves an LLM using tools to accomplish tasks autonomously, with human oversight at checkpoints. Both patterns are customizable and can be combined to fit different use cases.
Anthropic: Building effective agents
AI Summary
This piece describes two patterns for building LLM applications: evaluator-optimizer and agents. The evaluator-optimizer pattern involves an evaluator LLM providing feedback to an optimizer LLM, enabling iterative refinement. The agent pattern involves an LLM using tools to accomplish tasks autonomously, with human oversight at checkpoints. Both patterns are customizable and can be combined to fit different use cases.
Need help building compliant AI solutions?
AWS and HUMAIN announce a more than $5B investment to accelerate AI adoption in Saudi Arabia and globally
AI Summary
Amazon Web Services (AWS) and HUMAIN, Saudi Arabia’s AI innovation company, announced a $5 billion-plus strategic partnership to build an “AI Zone” in Saudi Arabia. The AI Zone will leverage AWS’s AI infrastructure, servers, and services to advance Saudi Arabia’s AI ambitions and support its Vision 2030. The collaboration aims to accelerate AI adoption across key sectors, foster a vibrant startup ecosystem, and develop AI talent in the Kingdom.
AWS and HUMAIN announce a more than $5B investment to accelerate AI adoption in Saudi Arabia and globally
AI Summary
Amazon Web Services (AWS) and HUMAIN, Saudi Arabia’s AI innovation company, announced a $5 billion-plus strategic partnership to build an “AI Zone” in Saudi Arabia. The AI Zone will leverage AWS’s AI infrastructure, servers, and services to advance Saudi Arabia’s AI ambitions and support its Vision 2030. The collaboration aims to accelerate AI adoption across key sectors, foster a vibrant startup ecosystem, and develop AI talent in the Kingdom.
AWS and HUMAIN announce a more than $5B investment to accelerate AI adoption in Saudi Arabia and globally
AI Summary
Amazon Web Services (AWS) and HUMAIN, Saudi Arabia’s AI innovation company, announced a $5 billion-plus strategic partnership to build an “AI Zone” in Saudi Arabia. The AI Zone will leverage AWS’s AI infrastructure, servers, and services to advance Saudi Arabia’s AI ambitions and support its Vision 2030. The collaboration aims to accelerate AI adoption across key sectors, foster a vibrant startup ecosystem, and develop AI talent in the Kingdom.
BBC: Netflix uses AI effects for first time to cut costs
AI Summary
Netflix used generative AI to create a building collapse scene in “The Eternauts,” completing the sequence ten times faster and at a lower cost than traditional methods. This marks the first time generative AI-created footage has appeared in a Netflix original series or film. The use of AI in entertainment is controversial, raising concerns about job displacement and content creation without consent.
BBC: Netflix uses AI effects for first time to cut costs
AI Summary
Netflix used generative AI to create a building collapse scene in “The Eternauts,” completing the sequence ten times faster and at a lower cost than traditional methods. This marks the first time generative AI-created footage has appeared in a Netflix original series or film. The use of AI in entertainment is controversial, raising concerns about job displacement and content creation without consent.
BBC: Netflix uses AI effects for first time to cut costs
AI Summary
Netflix used generative AI to create a building collapse scene in “The Eternauts,” completing the sequence ten times faster and at a lower cost than traditional methods. This marks the first time generative AI-created footage has appeared in a Netflix original series or film. The use of AI in entertainment is controversial, raising concerns about job displacement and content creation without consent.