Edition: July 10, 2026

Meta has shut down an Instagram feature that allowed users to create AI deepfakes of public accounts without permission, following significant backlash. The feature was part of Meta's new Muse Image AI model and was intended as a creative tool. However, it raised concerns over privacy and consent.

Apple has sued OpenAI, alleging that the AI startup's engineers stole Apple's hardware secrets to advance their own plans. The lawsuit names OpenAI, IO Products, and two former Apple employees, claiming a pattern of trade secret theft. The case involves high-profile figures, including Jony Ive, whose hardware startup was acquired by OpenAI in 2025.

Apple has filed a lawsuit against OpenAI, alleging trade secret theft and breach of contract, claiming that OpenAI's senior leadership, including former Apple employee Tang Tan, directed the misconduct. The lawsuit accuses OpenAI of using Apple's confidential information to develop its own hardware product, potentially competing with the iPhone. Apple is seeking to bar OpenAI from using its trade secrets and preserve evidence related to the case.

Apple is suing OpenAI for allegedly stealing trade secrets, including unreleased parts and prototypes, with the lawsuit accusing OpenAI's hardware chief of encouraging former Apple employees to bring confidential technology to the AI company. The lawsuit marks a significant escalation in the rivalry between the two tech giants. Apple claims OpenAI's hardware business is built on stolen secrets, while OpenAI denies any wrongdoing.

Hugging Face CEO Clem Delangue discusses the growing importance of open source AI, citing its cost-effectiveness and transparency as key benefits. Delangue also expresses concerns about the potential for big companies to control AI and highlights the need for open source models in areas like robotics. Hugging Face has become a hub for AI builders to share and download open models and datasets, used by roughly half of the Fortune 500.

SK Hynix, a South Korean memory chip giant, has raised $26.5 billion in its US market debut, the largest-ever US debut by a non-American company. The company's stock opened 14% above its IPO price, with demand reportedly over seven times the available shares. The funds will be used to build new facilities, including a fab in South Korea, and the US Commerce Secretary is urging the company to build new factories in the US.

Hugging Face CEO Clem Delangue discusses the rise of open source AI, citing cost savings as a key driver for companies to switch from rented AI models. The company's platform has become a hub for AI builders, with roughly half of the Fortune 500 using its open models and datasets. Delangue warns of the risks of big companies controlling AI development.

Sunrun, a solar and home energy storage company, is launching a pilot program to host AI data center compute units in customers' homes, offering compensation for participation. The company plans to sell the distributed compute power to enterprise buyers, such as AI companies. This approach aims to find new resources and space for AI compute infrastructure.

Anthropic has discovered a hidden space within its large language model Claude, revealing the model's thought process before generating responses. Meanwhile, OpenAI has unveiled its 'super app' ChatGPT Work, which combines its chatbot, coding tool, and new models. The app is designed to assist with work tasks and was launched alongside OpenAI's new GPT 5.6 models.

A new experiential gallery in Los Angeles, Dataland, is redefining AI art with its immersive exhibit 'Machine Dreams: Rainforest', which uses AI to create an interactive and dynamic environment. The gallery's founder, Refik Anadol, aims to prove that AI can be a tool for artistic engagement beyond generative 'slop'. The exhibit has welcomed over 10,000 visitors in its first two weeks, showcasing the potential of AI in the art world.

Nvidia CEO Jensen Huang has set a benchmark for engineers' AI token consumption, sparking a discussion on the trade-off between token budgets and personnel costs. Companies are increasingly allocating funds towards tokens, leading to layoffs, but a Gartner survey found that 80% of executives saw no correlation between headcount cuts and improved returns. By optimizing token usage through prompt caching and model selection, companies can reduce their token budgets without resorting to layoffs.

A European fertility group is calling for international limits on the number of children a single sperm donor can contribute to, citing concerns over donor-conceived people discovering numerous siblings and potential genetic risks. The proposal aims to establish a Europe-wide limit, with some countries already having laws in place to restrict donor contributions. The move seeks to protect the rights and well-being of donor-conceived individuals.

The UN AI for Good summit brought together representatives from the private and public sectors to discuss harnessing AI for humanity's benefit, with a focus on idealistic goals such as solving hunger, disease, and climate change. However, the conference was also marked by concerns over the risks of unchecked corporate monopolies and the erosion of human rights. The event highlighted the need for responsible AI deployment and greater access to AI models and compute resources for developing countries.

Meta has removed a controversial AI feature on Instagram that allowed users to modify photos from public accounts after facing backlash, citing that the feature 'missed the mark'. The feature was introduced earlier this week as part of a batch of new AI tools. Its removal comes amid concerns over potential misuse, particularly in generating explicit content.

Instagram head Adam Mosseri stated that users who don't like AI content shouldn't have it in their feed, but doesn't think AI content should be filtered out entirely. Instead, he suggests letting users know if content is AI-generated and allowing them to curate their own feeds. Mosseri believes users should have the option to have a feed with AI content if they choose to.

OpenAI has announced that its GPT 5.6 model will be the 'preferred model' for Microsoft's 365 Copilot, amid speculation about the status of their partnership. The move aims to bring advanced AI benefits to more users across Microsoft's productivity apps. OpenAI's software will continue to power Microsoft's apps, including Word, Excel, and PowerPoint.

Microsoft's carbon emissions increased by 25% in 2025, reaching 34 million metric tons, primarily due to the expansion of its datacenter infrastructure. This rise poses a challenge to the company's goal of becoming carbon negative by 2030. The increase is attributed to data center growth and changes in renewable energy purchasing practices.

The article discusses profiling attention mechanisms in PyTorch, focusing on understanding how different optimization techniques impact performance. It provides a step-by-step guide on how to profile a naive attention module and interpret the results. The profiling process involves tracing the execution of the attention module and analyzing the resulting trace to identify performance bottlenecks.