Hypatos GmbH

Hypatos GmbH is SOC 2 Type II compliant. On SOC2C this listing is Listed.

According to Hypatos GmbH's public trust center, Hypatos GmbH is ISO 27001 certified. On SOC2C this listing is Listed.

According to Hypatos GmbH's public trust center, Hypatos GmbH is ISO 27017 certified. On SOC2C this listing is Listed.

According to Hypatos GmbH's public trust center, Hypatos GmbH is ISO 27018 certified. On SOC2C this listing is Listed.

According to Hypatos GmbH's public trust center, Hypatos GmbH is HIPAA compliant. On SOC2C this listing is Listed.

Hypatos is a software company that specializes in advanced document processing and intelligent process automation. Our AI Agents leverage LLM combined with RAG, prompting and tooling to streamline repetitive, manual tasks, such as invoice processing, document classification, data extraction, master data matching. Our solutions help organizations achieve greater accuracy, operational efficiency, and cost savings.

_Data processed by third-party LLM Providers:_ We only utilise reputable LLM providers that have strict privacy standards in place - any data submitted via API to them is used solely to process the specific request and is not retained or used for training or improving their AI models. Hypatos is only using their base/pretrained models, which are stateless, meaning no prompts or output are stored in the models –after an output is generated, the prompt is immediately discarded. _Data processed and stored by Hypatos:_ By default, documents uploaded by customers to Hypatos Cloud are retained for 6 months. After this period, all data is automatically and permanently deleted. Documents used as a knowledge base may be retained for longer periods based on customer preferences, to ensure that the AI has access to sufficient customer-specific reference data so Hypatos can provide the highest quality and accuracy of services. Customer data is logically segregated between customers on the database layer and all locations where customer data is stored are encrypted using AES-256.

No. Our approach does not involve the conventional sense of “training” a model on customer data. Instead, we leverage pre-existing foundation large language models (LLMs) that are not retrained with customer-specific data. Our method uses in-context learning and advanced retrieval-augmented generation (RAG), which means that the model draws from relevant information provided during specific interactions without modifying the model itself or creating a customer-tailored generative model.

Hypatos hosts its cloud services on servers provided by Amazon Web Services (AWS). Customers can select between two AWS Regions for hosting their data based on data residency requirements: - eu-west-1 (Europe - Ireland) - us-east-1 (US - N.Virginia)

Hypatos commits to offer availability Service Level Agreements (SLAs) between 98% and 99.5% depending on the subscription plan selected by the customer.