Morning edition — Wednesday 4 November 2026

Apple released a Developer Preview that lets any app work with Siri AI on Tuesday, with the first iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas. Apps can expose their content and custom actions through App Intents, and new 'skills' let developers teach Siri AI to combine several actions for one request, such as checking a delivery or placing an order.

Apple says the feature ships to customers in a few months, and App Store submissions open in the spring. Siri AI is not available in the EU on iOS, iPadOS and watchOS; Apple says it is working with the EU to find a path forward.

On the same day, Apple's annual financial filing added warnings about AI agents. Apple wrote that agents 'can take actions autonomously, including interacting with internal or external systems that in some cases they may not be authorized to access, executing or altering code, sending communications, and completing transactions', and that they 'may behave in unintended or unpredictable ways'. Fortune reported that the filing says some consequences, such as deleting data or completing transactions, may be impossible to reverse.

OpenAI published a progress report on using its models for mechanical and electrical engineering on Tuesday. Its headline test: two staff members, working on it as a side project, used GPT-6 Astra to design a device for talking to Codex in under a week, aiming for no direct human edits. OpenAI says Astra designed the look, the enclosure, the circuit board and its components, the interface, and the firmware and software.

A contract manufacturer built a small batch of prototypes in under a month. OpenAI says getting the device ready for manufacturing still took several rounds of review, because Astra did not consistently account for the manufacturer's production limits, tolerances or assembly details.

The post also says model use across OpenAI's hardware teams, measured in tokens, has grown 23-fold since February. On BenchCAD 2.0, a test of reconstructing 3D parts and reading circuit-board connections, an unreleased internal model scored 0.875 on average against 0.594 for GPT-6 Astra. OpenAI is testing the work with Chip Ganassi Racing and has begun a research collaboration with Mazda on vehicle aerodynamics and crash safety.

AMD reported record third-quarter revenue of $13.4 billion, up 45% from a year earlier, after US markets closed on Tuesday. Its Data Center segment brought in a record $8.5 billion, up 97%, which the company put down to demand for its EPYC server processors and the launch of its Instinct MI450 Series GPUs.

For the fourth quarter, AMD expects revenue of about $17 billion, plus or minus 3%, which would be growth of about 66% on a year earlier. CNBC, citing LSEG, said analysts had expected $16.2 billion on average. Chief executive Lisa Su said AMD expects growth to accelerate as it scales production of its Helios rack-scale systems and 'Venice' processors.

The release says Anthropic, Meta, Microsoft, OpenAI and Oracle Cloud Infrastructure are running models and workloads on the first Helios systems. GAAP results include a $1.1 billion charge for the change in fair value of warrants that AMD has issued to certain customers, which moves with AMD's share price. Data centre sales made up about 64% of revenue, against 47% a year earlier, CNBC reported.

Some OpenAI employees raised concerns about the company's security practices even before ChatGPT launched in late 2022, and shared them with top leaders and the board, Reuters reported on Tuesday, citing five people familiar with the matter. Some came away believing management cared more about keeping pace with Google and Anthropic than about security, the people said.

According to the sources, anyone at OpenAI with access to its code could see technical details of many projects, because most code sat in a single repository. Three people said internal tools such as the sandboxes used to stress-test new models had not been regularly reviewed for security. OpenAI told Reuters its most sensitive work is held in separate repositories with strict access controls, that the main repository does not hold model weights, and that it now reviews internal tools regularly.

Two people said Thibault Sottiaux, a senior product executive, now leads an internal effort to secure OpenAI's systems and in August reassigned a quarter of production engineers to security work. Spokesperson Elana Widmann said OpenAI is one of the most targeted organisations in the world and invests heavily in security. The report follows the Justice Department's 13 October arrest of an OpenAI engineer charged with planting a recording device in an executive conference room.

Fal, a San Francisco start-up that hosts and runs AI video, image and audio models, has raised $1.3 billion in a Series F round led by Bessemer Venture Partners, with Nvidia, Sound Ventures and T. Rowe Price taking part, TechCrunch reported on Tuesday. The deal values Fal at $16 billion, double the $8 billion of its April round, according to Bloomberg's report cited by TechCrunch.

Fal announced its Series E and F rounds together and said they raised a combined $1.65 billion; Sequoia led the earlier round. Bloomberg reported that Fal passed $800 million in annualised revenue earlier this year.

A São Paulo court granted a preliminary injunction on Tuesday that bars Meta from collecting, copying, storing or using journalistic content from Brazilian media companies without authorisation, including to develop and train its AI systems, O Globo reported. The case was brought by three press associations: ANJ for newspapers, Abert for broadcasters and Aner for magazines.

Judge Guilherme Madeira Dezem found signs of copyright infringement and contractual abuse by Meta, according to O Globo. The order suspends part of Meta's new commercial terms for news outlets, which were due to take effect on 20 November. Meta must keep the outlets' accounts and tools on Facebook, Instagram, WhatsApp and Threads working, and may not penalise outlets that refuse the terms.

Meta faces a daily fine of R$1 million, capped at R$100 million, if it does not comply. O Globo notes that a Norwegian court temporarily suspended the same terms six days earlier, in a case brought by ten media groups.