What Anthropic's Q3 Numbers Mean for Vendor Risk

This Week in AI is an AI-generated weekly roundup, curated and reviewed by the Kursol team. We use AI tools to gather, summarise, and analyse the week's most important developments — then add our perspective on what it means for your business.

Anthropic's amended IPO filing put third-quarter revenue at about $17 billion and turned the company's operating result negative, and that is the number most vendor conversations will now revolve around. OpenAI also introduced a cybersecurity model this week that is aimed at finding and fixing software flaws. Our main stories this week are a follow-up on the Anthropic filing, the OpenAI release, and a US midterm issue that touches AI infrastructure. Two quick hits round out the edition.

Anthropic's Amended Filing Adds Q3 Numbers and Valuation Pressure

We covered the original filing in our October 27 breaking-news post, and our October 30 roundup looked at conflicting first-half figures. The filing has since been amended, on November 2, and the new section is the third quarter. Revenue rose to about $17 billion, up from $11.6 billion in the second quarter, and the company posted a $567 million operating loss after a $974 million operating profit in the prior quarter.

The bigger issue for buyers is the commitments. The same filing lists about $895 billion in commitments, nearly all for computing and data centres, with about $70 billion due within a year. A vendor carrying obligations that size has to keep growing revenue to meet them. The Wall Street Journal also reports that some investors want a lower valuation, in the range of $1.5 to $1.6 trillion, citing market volatility and AI-related risks. No share price or listing date has been set, so treat any timing claim as provisional.

Why it matters for your business: Vendor financial health is now a day-to-day procurement question. A large commitment base does not mean a vendor will fail, and it does not mean prices will rise. It does mean that pricing terms, service levels and data-export rights deserve a fresh look at renewal. Ask whether the contract locks in pricing for a set term, how much notice you get before a price or usage-limit change, and how you would move workflows to a second provider if service degraded. Teams that run an AI tool inside a core process should also be able to switch models without rebuilding the process. Our AI procurement checklist covers the wider list of renewal questions.

OpenAI's GPT-6 Cyber Targets Software Flaws, With Tight Access

OpenAI announced GPT-6 Cyber on November 5. It is a model built to find vulnerabilities, develop patches and check whether a fix holds. According to CyberScoop's report on the launch, OpenAI says the model found at least 81 previously unknown vulnerabilities in Firefox, developed exploits (working attacks) for 948 flaws in the Linux kernel, the core of the Linux operating system, and developed and validated a serious flaw in Brotli, a common data-compression tool, that could let an attacker run their own code on someone else's system. OpenAI says all of these were reported to the owners and patched.

Access is narrow. The first customers will be in OpenAI's Daybreak Red program, which covers advanced or nonpublic models for offensive security testing. OpenAI also describes added identity and account requirements, more frequent check-ins when a request involves high-risk behaviour, and monitoring for rule-breaking. The launch did not include a general release date.

Why it matters for your business: A tool that can find flaws can help a defender or an attacker, which is why access is being gated. For a company without a dedicated security team, the more practical effect is on the patch queue. As AI tools surface more problems, the bottleneck moves to whoever triages and applies fixes. Before your team adopts any AI security tool, confirm who on your side owns patch approval, how quickly high-severity fixes can ship, and whether the vendor's tool is allowed to test your own systems and under what controls. Our guide on how to tell whether your AI is still working describes the ongoing checks that catch problems after go-live.

Data Centre Backlash Becomes a Midterm Issue

Data centres were a central issue in Texas's midterm campaign, and Technology Review's November 4 analysis argues that the Democratic wins there could threaten the state's AI data centre boom. A companion piece on US energy policy after the midterms reports that voters largely backed restrictions on data centres in local ballot measures.

The story is less about one state than about where AI capacity gets built and what it costs. Local opposition has been driven by electricity prices, water use and land decisions, and those concerns cut across party lines. Expect permitting and grid rules to change in some areas, and expect the timelines for new capacity to shift with them. Results on the ground will vary, and this is a direction to watch rather than a settled outcome.

Why it matters for your business: Most companies will never choose a data centre site, but they will feel its effects through vendor pricing and availability. If a provider's capacity depends on approvals in a contested region, your service levels may depend on those approvals too. Ask your AI and cloud vendors where their compute runs, whether that location is exposed to local power or permitting rules, and what they would do to keep your service running if capacity were delayed. Build that answer into any cost model you use for AI automation. Our guide to calculating ROI on AI automation shows how to add infrastructure and usage costs to the business case.

Quick Hits: More AI News This Week

• OpenAI proposes that data centre developers cover their own power costs: OpenAI released a community compact and a compute campus policy blueprint. It proposes that developers cover electricity and supporting infrastructure costs for campuses they build and disclose water and power use. The proposal is a stated policy position, not a binding commitment.
• Australia's AI standards push has a committee deadline: The Australian government has said it will move from voluntary guidance toward a national AI standards framework, with supporting legislation expected in early 2027. The Joint Select Committee on Artificial Intelligence must report by 30 November 2026. This summary is from a late-August legal analysis, so check the government's own releases for the latest status.

What This Means for Your Business

This week's stories share one theme: the AI vendor is now part of your risk picture, and that picture includes its finances, its security posture and the power and permits behind its capacity. None of these is solved by a product demo. Each one shows up as a clause in a contract or a question you ask before signing.

Working through those questions is a large part of what our engineers do while embedded with a client. The usual first step is mapping which AI tools touch customer or operations data, which can take actions, and where each one runs, then deciding which dependencies the business can afford to lose. Our explainer on what a forward deployed engineer does describes how that work tends to run, and the free AI readiness assessment gives a quick view of where a team stands before it signs anything new.

The Bottom Line

The Anthropic numbers are a reminder that even well-funded AI companies have results that read differently depending on the line you look at. The OpenAI release shows that the most capable tools will come with access controls, and those controls will shape how fast a business can use them. The Texas story shows that the physical side of AI, including power and permits, can change what a vendor can deliver and when.

None of this is a reason to pause AI projects. It is a reason to write down what each tool can see, change, keep and depend on, and to review that list when a vendor publishes new numbers. The companies that do this now will find the next round of vendor news easier to read.

The gap between AI-ready and AI-late is widening every week. If you're unsure where your organisation stands, take our free AI readiness assessment to find out.

This Week in AI is Kursol's weekly analysis of the most important artificial intelligence developments — focused on what actually matters for your business. Subscribe to our RSS feed to never miss an edition.

FAQ