I

IamIP

The models power several core features of a patent platform that would still run without them. IamIP's product page describes a platform that monitors and analyzes patents filed worldwide, with automated alerts on competitors and technical areas, legal status watch, an idea database for invention decisions, analytics, collaboration and search with machine translation. The AI toolkit sits on top: natural language and similarity search, monitoring created from an AI search, an AI Patent Summarizer, an AI Claim Comparer and an AI Claim Clarifier.

A post of 16 July 2026 describes the search as embeddings with hybrid lexical search, reranking and retrieval augmented generation, and the pricing FAQ says every plan includes the Summarizer and the Claim Clarifier. Without the models, a patent database with monitoring, alerts and collaboration remains.

Grounding is described for the search and the claim definitions, and the accuracy figures carry no method. The post of 16 July 2026 describes hybrid search combining vector and lexical retrieval, a cross encoder reranker, and retrieval augmented generation in which the model answers from documents actually retrieved. It notes that embeddings are a statistical representation of meaning rather than understanding. The post of 21 September 2026 says the Claim Clarifier places definitions pulled from the patent's own description next to the claim they explain, and that search results are ranked by similarity score.

The same post says the Summarizer captures over 90% of key technical aspects and saves 15 minutes per document on average, and that 92% of users found the clarifier's definitions accurate. No test set, sample size or definition of a key aspect is given. Nothing says how often search misses relevant prior art or how a summary is checked against the full text.

The AI is presented as an aid to reviewers, and no oversight mechanism is described. The toolkit post of 21 September 2026 positions the Summarizer as helping reviewers assess relevance faster and the Claim Comparer as a side by side view, which keeps a person reading the patents. The search post of 16 July 2026 states a limit: embeddings do not provide true understanding or reasoning. Monitoring created from an AI search then runs continuously on that scope.

Nothing says what the AI may not be used for, whether a summary or comparison needs checking against the full text before a freedom to operate or infringement decision, or how a user learns that a monitoring scope built from an AI query is missing relevant filings. The platform is sold to management and R&D teams as well as to patent attorneys.

IamIP names companies in dated case studies, and the figures attached describe use rather than results. The case study page lists Bürkert (March 2026), Grundfos (July 2025, more than 100 employees across teams, with the AI Patent Summarizer said to save time), terpatent (October 2024), Georg Fischer (August 2024), Uhlmann Group (February 2024), TRUMPF (December 2023) and Sartorius (June 2023). IamIP's home page says more than 400 companies use IamIP and quotes reviewers from G2, GetApp and Capterra by first name and role.

The toolkit post's figures, 15 minutes saved per document and 92% of users finding definitions accurate, are not tied to a named customer or a method. The case studies are offered as downloads.

Security is asserted in general terms, and nothing addresses confidentiality, training or privilege. The patent attorney page says users work in a safe environment with advanced security measures and multiple user roles with varying access levels, and the AI search launch post of 1 July 2026 says all AI search processing and patent data handling occurs within IamIP's secure internal environment. No customer terms, data processing agreement or confidentiality commitment is published, and the privacy notice of 20 October 2024 covers the website, forms and webinars, not the platform.

Nothing says whether search queries, saved patents, collaboration notes or idea database entries, which can hold unpublished inventions, are used to improve models, who may access them, or how they are protected as confidential.

Nothing published addresses the advice line. IamIP's AI summarizes patents, compares claims across filings and clarifies claim terms, and its product page invites customers to secure their freedom to operate through automated monitoring. It is sold to management, R&D and IP teams as well as to patent attorneys and law firms. No terms of service, disclaimer, eligibility rule or statement that summaries and comparisons are not a legal opinion is published, so someone who is not a lawyer making a freedom to operate call from a claim comparison gets no caution.

No governance position on the AI is published. Neither IamIP's site nor its AI posts of 1 July, 16 July and 21 September 2026 name a person accountable for model behavior, describe testing before release, publish an evaluation method or results beyond user survey figures, or set out principles for how the AI is built and used. The search post describes the architecture and one limit of embeddings, which is engineering disclosure rather than governance.

Uneven output is a live question for a toolkit that compares claims across jurisdictions and languages and summarizes patents in many technical fields, and nothing addresses it.

IamIP publishes no data handling terms for the platform, only access roles and a statement that processing stays internal. The privacy notice, last modified 20 October 2024, covers website visitors, forms, chat, newsletters and webinars, names website tools such as Google Analytics, HubSpot, Salesforce and Intercom, and keeps contact form data until the person asks for deletion. It does not cover platform data. The patent attorney page mentions user roles with varying access levels, the AI search launch post says AI search processing stays within IamIP's internal environment, and a public status page reports service availability.

No retention period, deletion commitment, subprocessor list or breach notice for platform data, including idea database entries, is published.

Nothing published says who bears the loss when a search misses prior art or a summary is wrong. No terms of service, subscription agreement or warranty is linked from the site; the footer carries an imprint, a privacy notice and a cookie policy, and each plan's button leads to a quote form. No indemnity, cap or recourse is published for a firm relying on IamIP monitoring for a freedom to operate position.

No integration with practice systems is published. The product page describes alerts, collaboration, an idea database and search inside IamIP, and the partners page names a patent information group, a startup program, a German patent law firm and a business school, with no technical integration. No API, single sign on, Microsoft 365 connection, IP management system or docketing integration is named. A customer review on the home page mentions alerts by email.

Neither tenancy nor region is stated. The platform is used through a login at patents.iamip.com, and the AI search launch post says AI processing and patent data handling occur within IamIP's secure internal environment, without saying where that environment runs. IamIP's offices are in Stockholm, Cologne and Szczecin, which says nothing about hosting. The privacy notice, which covers only the website, refers to standard contractual clauses for transfers outside the EU by website tools. No hosting provider, data center country, tenancy model or region choice is published.

No independent security attestation is published. The product and patent attorney pages refer to advanced security measures and a safe environment, and the privacy notice mentions SSL or TLS encryption for the website. No ISO 27001 certificate, SOC 2 report, penetration test or trust center is named. The G2, GetApp and Capterra badges on the home page are review site awards, not security attestations, and the public status page reports availability only.

The architecture is described and no model or provider is named. The post of 16 July 2026, written by an IamIP backend developer, describes embeddings compared by cosine similarity, hybrid search combining vector retrieval with BM25 or TF IDF lexical search through reciprocal rank fusion, a cross encoder reranker, retrieval augmented generation, and fine tuning of both the embedding model and the language model. The AI search launch post says processing occurs within IamIP's secure internal environment.

Nothing names the embedding model, the language model, their providers or whether IamIP or a third party hosts them, and no notice of changes is committed.

Plans, inclusions and figures are published without a currency, and every plan leads to a quote form. IamIP's pricing page lists Starter at 0 per year, Pro at 89 per year and Enterprise at 299 per year, shown with no currency symbol or code. Plans set allowances of full and limited access users and monitorings, setup sessions of 45 or 60 minutes, and analytics in the higher plans, and an FAQ says every plan includes the AI Patent Summarizer and AI Claim Clarifier.

Extra monitorings and users can be added. The buttons under each plan lead to quote request forms rather than checkout. No currency, billing terms, contract length or implementation fee beyond the setup session is stated, so the actual cost needs a conversation with IamIP.

IamIP describes who it serves and what data it covers in substance, but not the limits. The site addresses management, R&D and IP teams and patent attorneys, small and medium businesses, enterprises and law firms, and names semiconductors, automotive, electronics and machinery, and aerospace and defense as industries. IamIP's pricing page states coverage of more than 105 countries, over 131 million patents and 338 million legal events, updated weekly, and the Claim Comparer works across jurisdictions and languages.

Not stated: which countries have full text, which have only machine translated text, how legal status data differs by office, and which languages the AI features support.