Claude

What Claude is

Claude is Anthropic’s AI assistant. You write a request in a chat window, or a business builds it into a product, and it replies in text. On this site it counts as a chatbot and as one example of generative AI, the wider category of systems that produce new content instead of only sorting or finding existing content.

The foundational courses cover it as one of the leading chatbots, alongside ChatGPT, Gemini and Copilot. It is not the tool any course is built on. It is one of the assistants a team should be able to place, so that people know what it is when a colleague mentions it.

The product and the Model

Like the others, Claude rests on a large language model. The model is the technology that predicts and generates language. Claude is the assistant built around it. That is why Claude, ChatGPT, Gemini and Copilot behave differently even though each rests on a language model. Each vendor trains its own, tunes it differently and wraps it in different features, and the wrapper decides much of what a business actually experiences.

What Hosted means

The distinction that matters most on this site is that Claude is hosted. It runs as a service on Anthropic’s infrastructure, and a business uses it by calling the model over an API or through a product built on one. There is no infrastructure to provision. Model updates and scaling are the provider’s job, not the customer’s.

The alternative is a self-hosted model, an open-weight one that an organisation runs on infrastructure it controls, on premises or in a private cloud. Nothing sent to that model leaves the environment. The cost is real. It needs GPU hardware sized to the model and the ongoing work of patching, scaling and monitoring, which in practice means in-house ML-ops capability. The local LLM against Claude review sets the two side by side. They solve different problems, and confusing “which is more compliant” with “which is more capable” is where the choice usually goes wrong.

What Anthropic states about your Data

Because the data goes to the provider, the terms of the account matter. The review reads Anthropic’s own privacy policy, which states that inputs and outputs from standard accounts may be used to train future models unless the user opts out. It also states that Business and Enterprise accounts are excluded from that default policy and are governed by their own customer agreement.

That difference is the practical point. Two people in the same company could use the same assistant under different terms, depending on which kind of account they signed in with. A business should read the agreement for the account it will actually use and decide what staff may enter, and it should not assume the consumer policy or the business one applies. The rule is drawn by the kind of data and the account, not by the brand name. Data protection covers what a commitment like this does and does not include.

Where the obligations attach

Running a model on your own hardware does not exempt it from the rules, and calling a hosted one does not put a business in breach. The EU AI Act and GDPR attach to what the system does and what data it processes, not to where the software runs. A poorly governed local deployment can breach either just as easily as a poorly governed hosted one.

The Act’s requirements scale with the risk classification of the system. A high-risk AI system still needs a risk management process and documentation whether it runs on a server down the corridor or behind an API. What changes between hosted and self-hosted is who is directly accountable for the data path and how much of it an organisation can see and control itself. The EU AI Act and GDPR pages set out what each one asks.

Hosted, Local or both

The review’s comparison is plain. Self-hosting trades ongoing infrastructure and ML-ops effort for direct control. A hosted model trades that control for capability and speed of deployment. Open-weight models are improving but are generally behind the frontier hosted ones, and a hosted service gives frontier-level capability without any infrastructure investment.

A business does not have to pick one. A common pattern routes sensitive or regulated work to a local deployment and less sensitive tasks to a hosted model. The right split depends on what is being processed and who is accountable for it. That is a question about the work, and it is why the review ends where it does. An audit’s Identify step is built to answer it, weighed against what a candidate process handles and not decided in the abstract.

How the Courses use it

Claude appears in the courses as a tool to compare and, in one advanced session, as a tool to work in. The Copilot and ChatGPT courses set the leading assistants side by side, Gemini, Claude, Perplexity, Grok, DeepSeek and Mistral among them, so a team can tell which job belongs to which tool and does not memorise one interface.

The advanced generative AI course goes further. Its spreadsheet block uses Gemini and Claude on Excel functionality, table improvement and a KPI dashboard. Its coding block covers Claude Code and Artifacts, alongside risks and mitigations and the importance of understanding what the tool produces. The same course opens with a decision framework on how to choose your tools and covers multi-tool work, which is the skill the comparison serves. That block includes multi-tool post creation and persona-based summaries, so participants see the same task done in more than one assistant.

The course also covers presentation creation, tools and connectors such as calendar communication and inbox triage, and research with a research assistant and a cross-checker. It closes with brainstorming to find your own use case. Claude is one of several tools across those sessions, and the point of using more than one is to learn which suits which job. The home page describes the business’s training in the same terms, as teaching teams on the AI tools they already pay for, with Claude named among them.

Mix-ups worth avoiding

Four confusions come up often enough to name. The first is treating the assistant and the model as one thing, which hides the fact that the wrapper decides much of what a business experiences. The second is assuming a hosted service is automatically the risky option, when the Act and GDPR attach to the system and the data and not to the location.

The third is the mirror image. A business that moves a workload onto its own hardware and considers the compliance question closed has not closed it, because the same obligations follow the system there. The fourth is assuming one policy covers every account. A standard account and a Business or Enterprise account sit under different terms, so the sign-in matters as much as the product.

Each of these comes from asking “which is better” before asking what the work handles and who is accountable for it. The review’s whole argument is that the second question comes first.

Where it goes wrong

Claude can produce an answer that reads well and is wrong. This is hallucination, and it is a property of how generative AI works, not a fault of one product. The courses treat it as a working habit. Ask for the source, read the source and treat the first reply as a draft.

Coding with an assistant adds its own risk. The advanced course covers it as risks and mitigations and puts the importance of understanding beside them. A tool that writes code quickly does not remove the need for a person to know what the code does before it is used. The same holds for a spreadsheet or a report. The larger the job handed over, the more the check matters.

Claude compared with other Assistants

The site does not rank the leading assistants, and the reason is stated in the courses. Which tool suits which job depends on the tools an organisation already has, so that is what sessions run on. Microsoft 365 Copilot sits inside Word, Excel, Outlook and Teams and answers from content the user can already see. ChatGPT is a separate destination with more room to build on. Claude sits with them as a hosted assistant reached on its own or through products built on it.

The useful question is not which is best in the abstract. It is which ecosystem a business already runs on, what the workflow needs, and which terms the account sits under. Gemini and ChatGPT have their own glossary pages, and the Mistral Le Chat and Perplexity pages cover two more of the assistants the courses compare.

Where it is taught and where it appears in an Audit

The comparison runs through several courses. The ChatGPT Enterprise course and Copilot Essentials both place Claude among the leading assistants. The advanced generative AI session uses it for spreadsheets and coding. Copilot Essentials is one of the six core AI courses, and the rest sit in the wider training catalogue.

An audit meets Claude as a deployment choice. Whether a process should call a hosted model or run a local one turns on what it handles and who is accountable for it, and the Identify step is built to answer that. Once a tool is chosen, training turns the decision into something the people using it know how to do.