Good morning, AI enthusiasts!

AI is moving deeper into the physical world and the apps we use every day. Today’s briefing covers OpenAI’s engineering experiments, Siri’s new developer opening and AMD’s data center surge, alongside fresh scrutiny of security and publishers’ rights.

Your AI Edge briefing:

THE QUICK SCOOP

OPENAI

OpenAI Takes AI Design From Screen to Physical Prototype source image

Summary: OpenAI says two employees used GPT-6 Astra to design a device for talking to Codex in under a week. The experiment shows progress across mechanical design, electronics and software, while highlighting the work still needed before manufacturing.

The Quick Scoop:

  • A contract manufacturer produced a small batch of prototypes in under a month.

  • The design needed several review cycles to address production tolerances and assembly requirements.

  • OpenAI is also testing engineering workflows with Mazda and Chip Ganassi Racing.

Your AI Edge: Try AI on a bounded engineering task, such as preparing drawings or exploring design alternatives. Keep qualified engineers responsible for verification and manufacturability.

APPLE

Apple Opens Siri AI to Every App Developer source image

Summary: Apple opened a developer preview that lets apps expose custom content and actions to Siri AI. Developers can build beyond predefined app categories and test more natural, context-aware interactions.

The Quick Scoop:

  • The preview runs in the iOS, iPadOS, macOS and visionOS 27.3 betas.

  • App Intents connects custom actions; new skills can combine several actions into one experience.

  • Apple describes this as early access, with customer shipping in a few months and App Store submissions opening in spring.

Your AI Edge: Choose one useful customer task and prototype it end to end. Test ambiguous requests and confirmation steps before letting an assistant place orders or change important information.

SPONSORED BY MAMMOTHCLUB

Mammoth Club Black Friday book bundle

Turn your next learning goal into a year of progress. Mammoth Club’s pre-Black Friday offer pairs 90% off annual membership with two free physical books and free international shipping.

  • Choose your books from a collection of 50+ titles.

  • Explore 3,000+ courses, 100+ ebooks and 1,000+ mock exams.

  • Get access to Mammoth Templates alongside your learning resources.

Book benefits require an active, paying annual membership. Free trials, non-annual and cancelled memberships are excluded. Black Friday signups receive two free books upon signup, then one free book each following year. Templates are not included in the free trial.

AMD

AMD’s Data Center Revenue Nearly Doubles source image

Summary: AMD reported record quarterly revenue of $13.4 billion, up 45% from a year earlier. Data center sales were the main driver as demand for EPYC processors and new Instinct accelerators strengthened.

The Quick Scoop:

  • Data center revenue reached $8.5 billion, up 97% year over year.

  • AMD attributed the segment’s growth to EPYC demand and the MI450 Series GPU launch.

  • GAAP net income was $2.4 billion; the separately reported non-GAAP figure was $3.5 billion.

Your AI Edge: The results reinforce the scale of AI infrastructure demand. When choosing compute, compare availability, software compatibility and cost for your own workload rather than relying on chip headlines.

EUROPEAN UNION

AI-generated conceptual illustration: EU Court Adviser Draws a Line Around AI News Reuse

Summary: An adviser to the EU’s top court said AI chatbots’ real-time retrieval and reuse of news may infringe publishers’ rights in some circumstances. The opinion concerns a dispute involving Google Gemini and Hungarian publisher Like Company.

The Quick Scoop:

  • The analysis distinguishes reported facts from a publisher’s protected expression.

  • It considers access restrictions and whether an answer reproduces recognisable expression that substitutes for the original article.

  • The opinion is nonbinding; the court’s judgment will come later.

Your AI Edge: For news-based AI products, pay attention to retrieval permissions and how much source expression appears in answers. The eventual judgment will matter more than treating this opinion as settled law.

OPENAI

AI-generated conceptual illustration: Reuters Examines OpenAI’s Internal Security Warnings

Summary: New Reuters reporting says employees raised concerns about OpenAI’s operational security over several years. OpenAI says it invests heavily in protection and has strengthened controls around sensitive work.

The Quick Scoop:

  • Reuters cited five people familiar with concerns about internal access and information sharing.

  • OpenAI says its most sensitive work uses separate repositories with strict controls, and its main repository does not contain model weights.

  • The company says it now conducts regular security reviews of internal tools.

Your AI Edge: As agents gain access to company systems, review permissions and internal tools alongside model quality. Broad access can create risks even when the model itself performs well.

SPONSORED BY MAMMOTHCLUB

Copilot Excel and prompting course illustration

Make Copilot part of your everyday workflow with Mammoth Club’s Complete Copilot Bundle with Excel and Prompt Engineering.

  • Explore and analyze Excel data using natural-language questions.

  • Practice prompts tailored to reports, emails and other Microsoft 365 tasks.

  • Work through downloadable practice files to put the lessons into action.

FAL

AI-generated conceptual illustration: Fal’s Reported $1.3 Billion Raise Targets More AI Compute

Summary: New reporting from Bloomberg, covered by Investing.com, says Fal raised $1.3 billion at a $16 billion valuation. The financing closed earlier, in October, and backs infrastructure for image, video and audio models.

The Quick Scoop:

  • Bessemer led the round, with investors including Nvidia and T. Rowe Price.

  • CEO Burkay Gur said rising demand had forced Fal to limit customer usage over the summer.

  • The company plans to add computing capacity and expand its team.

Your AI Edge: For media-generation workflows, evaluate rate limits and peak-time reliability as carefully as output quality. More funding signals expansion plans, but does not guarantee immediate capacity for every customer.

META

Brazilian Court Blocks Meta’s News-Training Requirement source image

Summary: A São Paulo court issued a preliminary injunction stopping Meta from collecting AI-training content from media outlets represented by three Brazilian press associations, AFP reports. The dispute concerns proposed business terms requiring free access to their work.

The Quick Scoop:

  • The judge suspended the disputed terms and barred penalties against outlets that reject them.

  • Noncompliance could bring a daily fine of one million reais.

  • The order is preliminary, and Meta has an opportunity to present its defence. It did not immediately comment to AFP.

Your AI Edge: Content rights remain a practical constraint on AI development. Track the scope and status of these disputes before assuming a platform’s proposed terms settle who can use publishers’ material.

THE MAMMOTH LOWDOWN

⚙️ TOOLS FOR AI EDGE

Five tools to explore for testing AI agents and understanding how they behave.

  • Opik - Trace agent steps and run evaluations across development and production.

  • Giskard - Test agents for security and quality issues, including prompt injection and data disclosure.

  • Inspect - Build AI evaluations with datasets, tasks, tools and scoring.

  • W&B Weave - Trace AI applications, compare prompts and evaluate outputs.

  • OpenLIT - Monitor AI agents with OpenTelemetry-based tracing, evaluations and cost tracking.

📢 What Else Is Buzzing In AI

OpenAI appoints Jill Hazelbaker - The former Uber executive will lead global communications and corporate positioning as Chief Corporate Affairs Officer.