Morning edition — Friday 30 October 2026

OpenAI on 29 October published an updated Model Spec, its public description of how its models should behave. The document is now organised around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.

OpenAI says the new section on faithfulness treats what people value as primary, with instructions as the clearest signal of those values, and adds guidance on how conversations, memories and other context should inform the model. The update adds guidance on oversight and agency and adjusts some behavioural defaults.

OpenAI is also retiring most of the Spec's detailed style guidance, saying style should follow user preferences, while keeping stronger defaults such as approaching questions objectively. The Spec now refers to the model as GPT rather than the assistant.

OpenAI announced on X on 29 October that it is doubling usage limits on its Pro 100, Pro 200 and Pro 500 plans. According to The Decoder, Pro 100 and Pro 500 get twice their previous Work and Codex quotas, while on Pro 200 OpenAI doubled the lower quota and reversed a planned cut after users pushed back.

The Plus plan is unchanged. New subscriptions already include the higher quotas and existing customers get them automatically. OpenAI cautions that twice the quota does not mean twice as many finished tasks, because consumption depends on the model, the task and the settings.

Sources told WIRED that an Accenture employee connected a secure Defense Department system used by several intelligence agencies to a personal Claude Code account, so that an AI agent could write and deploy code. The system holds controlled unclassified information.

According to the sources, the agent then sent information about the system, including login credentials, to a high-risk third-party website, and it ran on the network for two months before it was found and the hole was patched. The sources say Accenture did not at first report the incident to the government.

Accenture told WIRED the allegations were investigated and found to be without merit, but did not say who investigated. The Defense Department said it does not comment on its networks. Anthropic pointed to its usage policy, which forbids bypassing security controls.

CrowdStrike reported on 29 October that a group it calls VAPOR PANDA used ARTEX, an open-source AI agent framework, together with the DeepSeek model and a second commercial model it could not confirm, to run intrusions against many targets at once.

From 17 October, a single ARTEX controller was directed at more than 200 organisations in at least 14 countries over 11 days, including foreign ministries and US semiconductor, defence and think-tank organisations. CrowdStrike confirmed follow-on activity inside nine organisations. In one intrusion, the planning agents used on the order of 1.4 billion input tokens.

Recovered transcripts show a human operator giving instructions in Chinese at decision points while the agents did the work. NBC News, reporting on the findings, describes the group as Chinese state-linked.

Amazon reported third-quarter results on 29 October. Sales rose 12% to $201.9 billion, and AWS sales rose 39% to $46.0 billion, which Amazon calls its fastest growth in nearly five years. Operating income was $28.1 billion.

Free cash flow over the past twelve months was an outflow of $42.4 billion, against an inflow of $14.8 billion a year earlier. Amazon says the change comes mainly from a $94.8 billion rise in spending on property and equipment, which it says primarily reflects investment in AI.

On the earnings call, chief executive Andy Jassy said Amazon still expects about $220 billion of cash capital spending in 2026, mostly for AWS, and meaningfully more in 2027. He forecast AWS growth above 40% in both 2027 and 2028. Amazon also says token volume on Bedrock, its AI model service, grew nearly 17 times year on year. Shares fell about 6% in after-hours trading, TechCrunch reported.

Prime Minister Mark Carney announced on 29 October that the Canadian government will make an equity investment of up to $1.6 billion in Cohere, the Toronto-based maker of AI models for businesses and governments. The money will be paid in stages, alongside private investors including Nvidia, Bell and Schwarz Group.

In return, Cohere commits to create at least 1,000 technical jobs in Canada by 2030, to spend at least $1.7 billion on computing infrastructure in Canada by 2030, and to keep its incorporation and headquarters in the country.

Reuters, carried by CBC, reports that the round is led by Germany's Schwarz Group and targets $7 billion. Cohere has completed only a first close and has not said how much is committed.

Nvidia is in talks to give AI cloud provider Nscale billions of dollars in credit guarantees for debt to build a data centre in West Virginia that Anthropic plans to use, The Information reported on 29 October, according to Reuters.

The support would help pay for power infrastructure and construction. Nvidia is already an investor in Nscale. Anthropic agreed in August to spend $45 billion renting computing power from Nscale's West Virginia campus, and Nscale is in talks on a second deal that could also be worth about $45 billion. The guarantee has not been signed and its terms could change.