Top questions on AI & security
We’re on a mission to make it easy to use patent intelligence and to transform the processes that steal your time. AI has a huge role to play in helping us do that.
What type of AI do you use & what do you use it for?
Generative AI
Making patent content understandable
We use generative AI to summarise the patent content that is in public domain. Our AI Summaries are hosted on a separate in-house server, with access to a 3rd party API.
Machine learning: AI Classification
Create custom classifiers
Reflect your business and then our AI categorises patents using your own taxonomy.
Each model is specific to each client and no information is shared with other users of any Minesoft platform.
Labelled interactive images
Quickly understand a drawing
With labels based on the full text of the patent.
The automation uses publicly available patent information and does not rely on user prompts for input.
Quantity searching
Extract and search for quantities
No matter how quantities are referred to, ranges, values, metric, imperial, and all the various short cuts for describing them.
Our AI data mining technology is hosted on our in-house servers and uses pre-established data to find relevant results for all quantity searches.
Natural language processing: Intuitive search
Go beyond keywords and Boolean
Use natural language such as your invention to find the most relevant prior art.
To do this we use a vector-based search that finds relevant patents based on the free text that you input. Our in-house trained model uses vector mapping technology to convert your query and mathematically index patents.
All aspects of this search are performed in-house and are not shared with other users not used for training the model.
How do smart summaries work?
All our summaries are performed just-in-time via an automated third-party API request. No customer data is used at any time to generate the response.
Only the publicly available published patent information is shared with the 3rd party. The process is otherwise completely anonymized. No information specific to our clients is used for any purpose when summaries are requested.
Will the system learn from my data or my searches?
We don’t use customer data or queries to train models that are available to other users of any Minesoft product. The AI models are trained on the textual context of patents and the associated metadata. All necessary training for these AI models is done in advance to ensure the highest quality datasets and provide our clients peace of mind that none of their private information is being utilised to build our AI models.
Our AI classification system is hosted on our in-house servers, along with the user-defined taxonomy, tagging, and associated mode is proprietary to you. This data is stored using vector technology, so the text is encrypted, and no client data or info can be shared.
How do you train your model?
For models that are available to all users, we do not use client data to train/improve our AI. Our clients’ searches and the associated results are not shared outside of Minesoft and cannot be traced back to our customers.
For private models, like AI classification, the model learns from your behaviour and tagging. This information is entirely private and not sharable outside of your organisation.
How is input data for AI Search handled, used, and/or secured by Minesoft and/or any third parties?
There is no 3rd party access to AI Search inputs. The data from these searches are kept entirely on Minesoft’s in-house servers. Clients and users cannot see each other’s AI search data. Only a very limited number of specific senior personnel at Minesoft can access the information for support purposes, and only with prior written consent of the user in question.
No information regarding AI Search queries is ever used to initiate customer outreach and/or train our machine learning tools. No information specific to the user (name, employer, etc.) is used for any purpose when AI searches are run and/or stored. Machine learning training for AI Search is done independently of user queries altogether via a separate in-house process so no customer data is used at all for this.
Where are your servers/AI database hosted?
Our private cloud where we host our applications is in the UK and is bound by UK GDPR. There is an EU statute that deems UK GDPR to be equivalent to EUs data protection policy (Data protection adequacy for non-EU countries (europa.eu)). Allowing the free flow of personal data from the EU to the UK without additional safeguards.
How does Document Chat work?
Document chat is a function that evaluates the patent document’s claims and description and responds to your enquiries accordingly. We employ a third-party AI service provider to power this function, and all questions are sent to them anonymously. Neither Minesoft nor the third-party AI provider will learn or be trained based on your enquiries.
We advise against sharing any sensitive information in your input text, as it will be processed by a third-party AI service provider. While we strive for accuracy and relevance in this content generated by artificial intelligence, we cannot guarantee its reliability or completeness. We recommend that users thoroughly review the full patent documents and use their own judgement before making any decisions based on this content.
Do you have any type of certification?
Our applications are hosted in an ISO27001 certified data centre. ISO27001 is an international standard for information security management (ISMS). Our technologies are developed and deployed in accordance with relevant ethics guidelines and legal standards.
Further certification and policy documents are available on demand.
Who has access to my queries and saved histories? Where are they stored?
Only a very limited number of specific senior personnel at Minesoft can access the information for support purposes, and only with prior written consent of the user in question. The data from searches are kept entirely on Minesoft servers. Information used to populate searches are saved on a user’s search history and can be deleted at any time by the user themselves.
Is the data collected or the pretrained model used for the AI system(s) sourced from trusted sources?
Yes, we ensure that the data collected is received from trusted sources only. The data utilised to train the AI search model came from a public patent database. Minesoft Searching and analytics functionalities uses data information from our own in-house database.
How will you manage the unlikely event of disclosure of internal information?
In the unlikely event of an AI-related incident that impacts security or customer data, Minesoft will activate its incident response protocol to address and mitigate the incident. This includes timely communication with affected clients regarding the nature of the incident and corrective actions taken.
In general, we will follow the same process as reporting and responding to a cyber security incident. For more information, please access https://www.ncsc.gov.uk/collection/board-toolkit/planning-your-response-to-cyber-incidents. Further certification and policy documents are available on demand.