1. Headline & intro
When the people who actually build AI products change their daily tools, the market usually follows. At this year’s HumanX conference in San Francisco, the hallway talk wasn’t about OpenAI’s next big model or some flashy demo. It was about something more mundane and more important: which chatbot teams actually rely on to get work done.
According to reporting from TechCrunch, the answer, more often than not, was Claude. That apparently left ChatGPT oddly in the background, despite OpenAI’s colossal funding and looming IPO. In this piece we’ll unpack what that shift in buzz really means: for OpenAI, for Anthropic, for the rise of agentic AI – and for European companies trying to pick the right stack under tightening regulation.
2. The news in brief
As reported by TechCrunch, the HumanX AI conference filled San Francisco’s Moscone Center with thousands of industry attendees focused on one theme: agentic AI – automated systems that handle business and coding tasks with minimal human supervision.
On stage and on the expo floor, one product kept coming up: Claude, Anthropic’s chatbot and coding assistant. Multiple vendors told TechCrunch they and their teams now default to Claude, while some said ChatGPT and OpenAI had "fallen off" in quality and relevance. By contrast, ChatGPT drew noticeably fewer organic mentions in panels and conversations.
TechCrunch notes a growing perception that OpenAI has lost strategic focus, after shelving side projects like the Sora video generator and an ill‑fated "sexy" ChatGPT, and after controversies ranging from political work with the Trump administration to injecting ads into ChatGPT. OpenAI board chair Bret Taylor publicly defended CEO Sam Altman during the event.
Meanwhile, OpenAI announced a new $100/month ChatGPT tier aimed at heavy coding use, clearly positioned against Anthropic’s Claude Code. Citing Wall Street Journal analysis, TechCrunch adds that OpenAI and Anthropic are both among the fastest‑growing tech businesses ever, with Anthropic apparently catching up fast among enterprise users.
3. Why this matters
HumanX is not a mass‑market consumer show. It’s where the architects of next‑generation products – CTOs, staff engineers, founders, product leaders – compare notes. When that crowd starts saying "we’ve moved to Claude for work," it’s an early warning that power‑user sentiment is shifting.
Those people decide which API goes into a new SaaS platform, which agent framework becomes the default for internal tooling, which coding assistant is rolled out to thousands of developers. Their preferences don’t just shape usage numbers; they lock in ecosystems for years.
Anthropic’s advantage here isn’t only model quality. It’s positioning. The company has spent years branding itself as the safety‑obsessed, research‑driven alternative: long context windows, more predictable behavior, and explicit "constitutional" guardrails. For teams shipping agentic systems that can execute code and touch production data, perceived reliability and guardrails matter more than viral demos.
OpenAI, by contrast, is suffering from narrative drift. The pivot away from headline‑grabbing experiments like Sora toward more focused business and coding tools is rational, but from the outside it can look like backtracking under pressure. Add political controversy and the deeply unpopular move to put ads into ChatGPT, and you get a sense that the company is reacting to events rather than defining them.
The immediate winners are:
- Anthropic, gaining mindshare among practitioners.
- Enterprises, which now have at least two credible, hyperscale providers instead of one quasi‑monopoly.
- Tooling and infra players, who can build multi‑model workflows and arbitrage strengths.
The losers are smaller model vendors and niche coding tools, which now face not one, but two giants determined to own the agentic AI stack.
4. The bigger picture
The Claude vs. ChatGPT buzz at HumanX sits inside a much larger trend: the shift from chatbots to agents.
In 2023–2024, the big story was conversational AI: chat interfaces, summarisation, image generation. Today, the frontier is systems that take actions – writing and running code, moving data between SaaS tools, triaging tickets, drafting and sending emails, even making limited financial decisions within guardrails.
Whoever becomes the default engine for those agents wins something far more durable than consumer attention: they win a place deep inside business workflows. Once your CI/CD pipeline, your CRM automations or your customer‑support triage run on a given model, switching is non‑trivial. That’s exactly how AWS, Salesforce or Microsoft Office became so entrenched.
Historically, these markets tend to settle into a duopoly or triopoly. Think iOS vs. Android, or Chrome vs. Safari/Edge. For a while, OpenAI looked poised to be the Google Chrome of LLMs – so dominant that everything else was a niche. The mood at HumanX suggests a different outcome: Anthropic has secured its place as a co‑equal pillar of the stack.
This doesn’t mean other players vanish. Google is pushing its own agentic vision through Gemini and Workspace; Meta continues to drive open‑source‑first strategies; open models such as Llama are quickly getting good enough for many workloads. But in the specific market for general‑purpose, hosted agents used by enterprises, the race currently looks like OpenAI vs. Anthropic.
And here’s the twist: both are, according to TechCrunch’s references to WSJ data, among the fastest‑growing companies in tech history. "Falling off" for OpenAI doesn’t mean collapsing. It means losing the aura of inevitability – and that changes how developers, enterprises and regulators negotiate with it.
5. The European / regional angle
From a European perspective, this emerging duopoly lands just as the EU AI Act starts to bite. European organisations will soon have to document not only what they built, but which model they used, how they evaluated risks, and what mitigation measures exist.
That plays directly into Anthropic’s narrative. A vendor that can show well‑documented alignment methods, constitutional principles and conservative deployment choices is easier to defend to a regulator or a works council than one making headlines for political entanglements or ad‑driven consumer products.
At the same time, both OpenAI and Anthropic remain U.S.‑based. Cross‑border data transfers, GDPR compliance, and Digital Services Act obligations still apply, whatever the marketing says. European contenders such as Mistral AI (France) and Aleph Alpha (Germany) are already pitching themselves as "sovereign" alternatives, and the DACH region in particular is receptive to that message.
What HumanX reveals, however, is that European developers will not compromise on capability. A startup in Berlin or Ljubljana building an AI agent to manage cloud infrastructure will pick the model that debugs fastest and fails least, then wrap it in encryption, audit logs and contractual protections. If that’s Claude today and something else tomorrow, so be it.
For EU enterprises, the pragmatic answer is a multi‑model strategy: combine OpenAI, Anthropic and strong European or open‑source models behind a routing and governance layer. The provider that makes this easiest – with clear logging, eval tools, EU data regions and contractual guarantees aligned with GDPR and the AI Act – will have an edge.
6. Looking ahead
Conferences like HumanX are often leading indicators. If that holds, expect the next 12–24 months to look roughly like this:
- Anthropic deepens its enterprise beachhead. More teams standardise on Claude for agentic coding and back‑office automation, especially where hallucinations or policy violations are unacceptable.
- OpenAI doubles down on monetisation. The $100 ChatGPT tier is likely the start of a broader push: richer coding tools, better team management, maybe vertical offers for finance, healthcare or customer support.
- Agent tooling matures. Frameworks for orchestrating, testing and monitoring agents will become as standard as CI/CD pipelines. That reduces switching costs and makes "best model per task" architectures more common.
- Regulation becomes productised. Expect audit dashboards, model cards, logging APIs and compliance reports to move from PDFs to real‑time product features – especially for EU customers.
The open questions are non‑trivial. How far will enterprises let agents operate autonomously before legal departments push back? Will OpenAI’s upcoming IPO lock it into aggressive growth targets that further alienate safety‑conscious users? Can Anthropic find enough capital and distribution to compete without compromising its safety‑first message?
For European founders and CIOs, the opportunity is clear: build the governance, integration and domain expertise layer on top of these models. The risk is sleepwalking into single‑vendor dependency, only to discover that your entire operations stack is optimised for one provider’s roadmap.
7. The bottom line
The buzz around Claude at HumanX doesn’t mean OpenAI is finished; it means the age of AI monoculture is over. Power users are voting with their workloads, and Anthropic has earned a seat at the top table of enterprise AI.
For businesses – especially in Europe’s tightly regulated environment – this is good news. Competition should mean better tools, clearer governance features and more leverage in negotiations. The real question is not "Claude or ChatGPT?" but rather: what’s your strategy when the AI tool your team swears by today stops being the obvious choice tomorrow?
