The one thing
Free AI from Z.ai can hack unpatched Firefox for about $1,000
Why it mattersAttackers can now turn a published fix into a break-in for about $1,000, so any delay in installing browser updates is a real exposure.
Seventy: Friday, 30 October 2026
What happened. Epoch AI reports that Z.ai's openly available GLM 5.3 turned published-but-undelivered Firefox security fixes into a working attack that stole a file from a visitor's computer with no clicks. Each Firefox build from mid-September to late October fell to it, at roughly $1,000 per build; Mozilla's 159.0 update on October 27 closes one flaw the attack relied on.
Teams at your company can now share unused AI computing power on Amazon's cloud
Why it mattersCompanies running several AI teams on shared GPUs on AWS can pool idle capacity instead of administrators negotiating access by hand.
What happened. AWS Batch now has quota management for Amazon SageMaker Training jobs, letting teams lend and borrow shared GPU capacity. It builds on AWS Batch's existing queuing, scheduling and retry support for ML training, aimed at cutting the coordination work administrators face as more teams onboard.
03Safety
OpenAI rewrites the rules for how its AI is meant to behave
Why it mattersIf you build on OpenAI's AI, its behaviour rules just changed, so re-test your customer-facing prompts and workflows against the new version.
What happened. OpenAI revised its Model Spec, the document describing how its AI systems are meant to behave, and organised it around five core values. Among them are good judgment, keeping people in control, and staying true to what users actually intend.
04Models
NVIDIA's AI that reads text, images and video can now run on AWS
Why it mattersTeams already running on AWS Kubernetes can host one NVIDIA model for both text and visual work, instead of running separate systems.
What happened. NVIDIA's Nemotron 3.5 Super VL, a 120B-parameter model that activates 12B per token, is now available on AWS. It handles text, images, documents and video, and AWS has published steps to deploy it on Amazon EKS and SageMaker HyperPod EKS.
AWS now lets you keep and search the full history of your automated workflows
Why it mattersTeams running multi-step workflows on Lambda can now keep execution history past the standard retention period and query it in plain language.
What happened. AWS Lambda has released a plugin, the Durable Execution Telemetry Exporter, that records each stage of a durable function run and sends it to destinations you pick. AWS also released an MCP server so AI assistants such as Kiro can query that data in plain language.
06Funding
Microsoft borrows $25 billion by selling bonds to investors
What happenedMicrosoft priced a $25 billion sale of unsecured bonds on October 29, 2026, split into floating-rate and fixed-rate portions maturing in November 2028 and 2029. The fixed-rate pieces include $2.00 billion at 4.85% and $3.25 billion at 5.00%.
That's the whole issue.
6 stories, about 2 min 18 sec. Nothing was left out to fill the time.
Words of the day
- Model Spec
- OpenAI's public document setting out how its AI systems are intended to behave and which principles they should follow.
- SageMaker HyperPod
- An AWS service for running large GPU clusters for model training, with automatic repair of failed nodes.
- MCP
- Model Context Protocol: a common standard for connecting AI assistants to tools and data — your files, your database, your ticketing system — without writing a bespoke integration for each one.
“High expectations are the key to everything.”