The EU AI Act – Supporting innovation and building trust across the financial services industry.

The EU AI Act – Supporting innovation and building trust across the financial services industry

The EU AI Act was agreed by the European parliament in December 2023 and the financial text is likely to be published in early 2024.

This act will apply to all industries across the European Union and is aimed at continuing to foster innovation while ensuring the protection of individual’s rights, through stricter regulation of high-risk AI technologies and the promotion of transparency and trust across AI technologies.

It recognises that innovation is essential for competitiveness, so this Act also includes the creation of regulatory sandboxes to facilitate the development, testing, and validation of innovative AI systems under strict regulatory oversight.

The Act establishes rules and obligations for AI technology, based on the potential risk to the user and society.  Five risk levels are defined, with stricter obligations for technologies deemed to be at higher risk.

Technologies with an unacceptable risk are banned, such as systems which aimed at exploiting vulnerabilities or behavioural manipulation. Technologies which are deemed as high risk, with the potential to impact on fundamental rights, democracy and health and safety, will be required to comply with extensive governance activities to ensure these technologies are compliant with the Act.

AI systems which are categorised as limited risk, will need to ensure that they are fully transparent, this means for example, that users should be aware if they are interacting with an AI chatbot or a human.

What does the EU AI Act mean for financial services?

Many of the AI technologies used across the financial services industry fall into the high-risk category, such as trading algorithms, risk analysis and credit scoring.  The onus will be on financial institutions to demonstrate that their models can be understood and that their underlying data is unbiased and of good quality. 

Matching the requirements of the EU AI Act has the advantage of winning consumer trust, as consumers are becoming very aware of the importance of ethical AI and the need to respect their rights and privacy.

While the EU AI ACT might take two years before coming into force, financial institutions should act now to analyse their AI technologies and make any necessary changes to comply with the required obligations.


Lingua Custodia’s Generative AI Document Analyser


Our latest generative ai financial document processing technology, our document analyser allows the rapid extraction of key data from large pdf documents such as the EU AI Act. It’s fully secure, like our other technologies, multilingual, and provides the source referencing!

You can test it here!



Generative AI for finance

Generative AI for finance


Generative AI is a powerful innovation and one which is being rapidly adopted by the financial industry to improve productivity and enhance workflows. Machine learning and AI have been used over the last decade within the financial services industry to automate and enhance previously manual processes such as fraud detection and compliance.  However, the usage of generative AI for finance can add even more value and be used across a number of key areas, providing a clear competitive advantage

What is Generative AI?

Generative AI is a type of AI which can generate content.  It can combine data from large language models (LLMs) and algorithms to generate content based on patterns it observes in other content.

It’s ability to generate content takes it beyond traditional machine learning.  Machine learning focuses on recognising patterns and making predictions and decisions based on this data.  Generative AI is able to generate new content based on the data is was trained on, so creating new content which follows the underlying data patterns.

What are the use cases for Generative AI for finance?

There are several use cases for Generative AI for finance. 

Fraud

Generative AI can be used to spot patterns and identify anomalies, such as transactions which do not follow typical patterns, which can then be flagged for further investigation.  This helps to improve the productivity and efficiency of the fraud team who can focus on a specific subset of transactions.

Improve Client Satisfaction

Generative AI can be used to in chat bots to provide human like interactions, to help answer customer queries and questions, for example responding to queries relating to balances and transactions.  The customer experience can also be personalised with the analysis of client data to provide recommendations for specific products.

Data Analysis and Extraction

As many financial institutions have to deal with large volumes of data which is time consuming and laborious, Generative AI can be used to rapidly summarise large documents, extracting the key information and providing summarises for further review.

Lingua Custodia’s document analyser was created to meet the demands of its clients for a technology able to rapidly summarise and extract key data from large documents.  It can also be used in conjunction with its specialised machine translation engines, to extract the data in different languages.  This allows its clients to query large documents in a different language to one the document is written in, which is invaluable for international financial institutions. 

Innovation in Finance – Fintech Campus Corporate with ING

innovation in finance

Our Managing Director Frederic Moioli was delighted to take part in this campus at the The LHoFT – Luxembourg House of Financial Technology in partnership with ING Luxembourg, which was very focused on driving innovation in finance, with lively and interesting discussions on key trends for the financial sector. Lingua Custodia with its extensive expertise in financial document processing has always been aware of the importance of innovation in finance to optimise workflows and add value for its financial clients.

The day consisted of a series of key discussions on current innovations in finance and finished with an interactive workshop to provide attendees with an opportunity to explore innovative technologies.

Lingua Custodia, represented by Frederic, participated in a session on the future of finance through the application of Generative AI in financial services.  Generative AI is currently being used to automate financial tasks, rapidly analyse and summarise key data as well as improve client satisfaction through the creation of chatbots.

Frederic was very happy to share his knowledge and experience with the attendees, and highlight how Lingua Custodia has been driving innovation in finance since its inception. 

Size no longer matters for large language models!

Lingua Custodia’s new compact open source language model Fin-Pythia-1.4B outperforms larger language models.

The French Fintech company Lingua Custodia, a specialist in Natural Language Processing (NLP) applied to Finance since 2011, releases its first open source language model on the Hugging Face Hub specifically trained for sentiment analysis of financial text.

Fin-Pythia-1.4B is a language model that’s been fine-tuned on financial documents and instructions. It can understand complex financial jargon and terminology. It is compact in size, which makes it fast to run without compromising the quality of the output.

Lingua Custodia’s open source language model is extremely accurate in analyzing financial sentiment and outperforms well-known models like GPT-4 and BloombergGPT. 

Raheel Qader, The head of Lingua Custodia’s Research and Development lab highlights “to produce accurate language models, the essential bases are data, expertise and experience.  Following 4 months of research, we were delighted to find that our open source language model outperformed both GPT-4 and and BloombergGPT in various financial NLP tasks.  This demonstrates clearly that size is not necessary to create powerful models and I am delighted that the research team here at Lingua Custodia was able to bring the model to the open source community so rapidly.  This places Lingua Custodia firmly at the forefront of generative ai technologies“

Fin-Pythia-1.4B model card is available at this link

https://huggingface.co/LinguaCustodia/fin-pythia-1.4b

Lingua Custodia will be releasing a series of other large language models targeting various financial document processing use cases as part of its research and development strategy for 2024.

New Study Reveals AI’s Carbon Footprint is Much Lower Than Humans’

New Study Reveals AI’s Carbon Footprint is Much Lower Than Humans’

A new study published in Social Science Research Network found that AI systems have a much lower carbon footprint than humans for tasks such as writing and creating illustrations.

The researchers compared the emissions generated by AI systems like ChatGPT and DALL-E to the emissions generated by an average human completing the same tasks. They found that the AI systems emitted 130 to 1500 times less CO2 per page of text and 310 to 2900 times less per image. Researchers calculated the carbon footprint of the AI systems by looking at the emissions from training the models and from generating each individual response. Even when factoring in energy used for model training, the AI systems were far more efficient than humans. The emissions generated by a laptop or desktop computer used by a human were also greater than the AI systems. The researchers note that AI could play an important role in reducing emissions for certain activities. However, they stress that AI does not substitute for all human tasks and that factors like job displacement must be considered. They suggest collaboration between AI and humans as the best approach in many fields.

While AI emissions may grow as the technology advances, this study highlights an important benefit of AI systems as they stand today. It adds a new perspective to concerns about the carbon footprint of AI by comparing it directly to human activities. Even with current technology, AI enables the completion of common tasks at much lower emissions than humans can achieve.

Lingua Custodia’s Generative AI Document Analyser

Lingua Custodia’s Document Analyser tool for rapid data extraction!

Our new data extraction technology, the Document Analyser, is easy to use and allows swift extraction of key information from large documents. It was developed inhouse by our machine language experts and is fully secure, like all our other financial document processing technologies. The use cases include compliance and due diligence queries, calls for tender and fund and research queries.

The Document Analyser is easy to use. You simply upload your document and begin to type in your questions. It is available in a multilingual format, which means that you can load the document in one language and ask questions in a different language.

Do not hesitate to try it! You can sign up for a 14 day test access which allows you to try our AI translation services and the Document Analyser.

Lingua Custodia proves conclusively that it is possible to develop cutting edge generative AI tools without a 10M€+ investment!

The French Fintech company Lingua Custodia, a specialist in Natural Language Processing (NLP) applied to Finance since 2011, releases its new Document Analyser, a generative AI tool allowing its 10,000+ financial users to easily and securely answer due diligence questionnaires and find information in large volumes of documentation.


With one click and the input of questions, the users of Verto, Lingua Custodia’s secure document processing platform, can now retrieve information in multiple languages from vast amounts of data. The aim of the document analyser is to meet the business case for speed and accuracy in responding to regulatory, client and investment queries.


The Lingua Custodia team of research scientists and engineers had been working on the product design for the Document Analyser tool for over a year to propose this secure solution hosted in the EU on secure and segregated GPU servers.


Olivier Debeugny CEO of Lingua Custodia declared : “Generative AI tools are hot topics at the moment for funding rounds, I however firmly believe that innovative document processing technologies can be produced without large investment, provided you have a lot of data, experience and a skilled team! Lingua Custodia has been at the forefront of language technologies since its creation in 2011 and we believe the Document Analyser will empower our clients, improving operational efficiency, and enhancing their decision-making processes.”


The Document Analyser is available on Lingua Custodia’s financial document processing secure platform, together with its translation, transcription and other data extraction AI tools.

IA Frugale : Sans lever des millions d’euros, Lingua Custodia met à disposition de ses clients un nouvel outil d’IA générative

Sans lever des millions, Lingua Custodia met à la disposition des professionnels de la finance un nouvel outil d’IA générative répondant à des cas d’usage précis

Lingua Custodia, la fintech spécialisée dans les technologies du langage depuis 2011, a mis en ligne sur sa plateforme Verto accessible par plus de 10 000 professionnels de la finance, un nouvel outil permettant d’interroger des documents confidentiels multilingues.


C’est en analysant les besoins de ses clients, en utilisant sa longue expertise en technologie du langage et ses corpus linguistiques développés depuis plus de dix ans que l’équipe d’experts de Lingua Custodia sort aujourd’hui sur le marché un nouvel outil d’analyse de documents sécurisé et multilingue spécialisé pour le secteur financier.


Ce nouvel outil permet par exemple d’interroger des procédures internes pour répondre très rapidement à des questionnaires reçus de régulateurs ou clients, des situations très fréquentes dans les institutions financières.


Olivier Debeugny, Président de Lingua Custodia, déclare : « Je suis particulièrement fier de notre équipe qui illustre un autre angle de l’Intelligence Artificielle frugale. Avec des bases de données de grande qualité constituées sur plusieurs années, notre équipe de chercheurs très expérimentés travaillant main dans la main avec des universités et laboratoires de recherche, et une compréhension des besoins de nos clients, nous avons su mettre en ligne, sans devoir faire appel à des financements massifs, un nouvel outil à la pointe de l’état de l’art apportant une réelle valeur ajoutée à nos clients. »


L’analyseur de document est disponible sur Verto, la plateforme de traitement documentaire sécurisée de Lingua Custodia et vient s’ajouter aux outils de traduction automatique, de transcription et autres outils d’extraction de données.

A propos de Lingua Custodia


Lingua Custodia est une Fintech leader du Traitement Automatique des Langues (TAL) pour la Finance basée en France et au Luxembourg. Elle a développé son expertise avec une offre pointue de traduction automatique spécialisée par type de document financier. La société propose aujourd’hui également des services de transcription automatique, des services d’analyse linguistique de documents et des services d’extraction de données via sa plateforme en ligne ou par API. Ses clients sont des institutions financières et les départements financiers de grandes sociétés et ETI.


Contact Presse
Charlotte BAIN , CAO : charlotte.bain@linguacustodia.com / +33 1 83 43 95 25

Do Multilingual Language Models Think Better in English?

Do Multilingual Language Models think better in English?!

Most large language model (LLM)-based chatbots are trained on data from dozens of languages, but English is still the dominant language, as most of the web data is written in this language. Because of this, multilingual LLMs have much better understanding and generation capabilities in English than the other languages. However, we still need LLMs to perform well in other languages to ensure accurate and reliable results in multilingual contexts.

One solution which has been around for a while is to detect the input language, translate it to English, let the LLM process the question and generate the answer in English. Finally we can translate the response back to the desired language of the user. Although this process works well, in practice, it means you will need a few extra tools, such as a language detector and a machine translator. These additional tools can add to the complexity of the project.

In an experiment conducted by the University of the Basque Country, researchers confirmed that multilingual LLMs perform better in English than other languages seen during training. Their study shows that using multilingual LLMs to translate the input into English and perform the intended task over the translated input works better than using the original non-English input. This work shows that letting the model translate the input by itself can achieve almost the same performance as using an external translation system. This opens up opportunities for end-to-end multilingual chatbots and other generative AI models without relying on external translation systems. 

Lingua Custodia’s VERTO NLP financial document processing platform allows you to easily translate documents from several languages to English.  Our NLP translation service detects both the source language and document type which helps to optimise the translation quality. Our generative AI document analyser service then allows you to process questions in English, and again the responses can then quickly be translated back to the necessary language.  

How LLMs (Large Language Models) use long contexts

Large language models (LLMs) are capable of using large contexts, sometimes hundreds of thousands of tokens. OpenAIs GPT-4 is capable of handling inputs of up to 32K tokens, while Anthropic’s Claude AI can handle 100K context tokens. This enables LLMs to treat very large documents which can be very useful for question answering or information retrieval.

A newly released paper by Stanford University examines the usage of context in large language models, particularly long contexts for two key tasks: multi-document question answering and key-value retrieval. Their findings show that the best performance is typically achieved when relevant information occurs at the beginning or end of the input context. However, the performance of models significantly declines when they need to access relevant information in the middle of long contexts.This could be attributed to the way humans write, where the beginning and concluding segments of text mainly contain the most crucial information.

These findings show that one needs to be careful when using LLMs for search and information retrieval in long documents. Information found in the middle might be ignored by the LLM and hence wrong or less accurate responses will be provided.

Lingua Custodia has over 10 years of experience in language technologies for financial document processing and we are very aware of the importance of context for search and information retrieval sentiment analysis, content summary and extraction. We continuously study the impact of context size of these language models

Our expert team consists of scientists, engineers and developers, so we are well placed to create, customise and design secure LLMs which are perfectly tailored to meet your business needs.