Study: RAG-based LLMs less safe than non-RAG
Researchers at Bloomberg have found that retrieval-augmented generation is not as safe as once thought. As a result, they put forward a new taxonomy to help firms mitigate AI risk.

In the world of large language models, the retrieval-augmented generation (RAG) framework has gained favor as companies have looked to develop LLMs. RAG approaches grew in popularity about a year ago as the framework was viewed as a way to reduce hallucinations related to generative AI. IBM wrote in August 2023 that RAG LLMs have access to the most current, reliable facts.
Broadly, RAG-based LLMs have been deemed safer from the danger of answering “harmful questions” than non-RAG LLMs—but a new
Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.
To access these options, along with all other subscription benefits, please contact info@waterstechnology.com or view our subscription options here: http://subscriptions.waterstechnology.com/subscribe
You are currently unable to print this content. Please contact info@waterstechnology.com to find out more.
You are currently unable to copy this content. Please contact info@waterstechnology.com to find out more.
Copyright Infopro Digital Limited. All rights reserved.
As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (point 2.4), printing is limited to a single copy.
If you would like to purchase additional rights please email info@waterstechnology.com
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (clause 2.4), an Authorised User may only make one copy of the materials for their own personal use. You must also comply with the restrictions in clause 2.5.
If you would like to purchase additional rights please email info@waterstechnology.com
More on Emerging Technologies
Clearstream and SimCorp team up, Bloomberg goes dark, TT expands FX coverage and more
The Waters Cooler: Google’s AI mishaps, a new liquidity source for ASX customers, and new integrations can be found in this week’s news roundup.
Bloomberg’s chatbots streamline external data sharing, OTC trading
The vendor details the expansion of chatbot support outside of a firm’s enterprise, and addresses user reservations about cost and spambots.
Symphony introduces Cloud9 mobile app
The messaging and collaboration platform is extending the “turret-like” technology and connectivity of Cloud9 to smartphones.
EU banks want the cloud closer to home amid tariff wars
Fears over US executive orders have prompted new approaches to critical third-party risk management.
Growing pains: Why good data and fortitude are crucial for banks’ tech projects
The IMD Wrap: Max examines recent WatersTechnology deep dives into long-term technology projects at several firms and the role data plays in those efforts.
Waters Wavelength Ep. 317: Bitdefender and Transilvania Quantum
This week, Bitdefender’s Adrian Coleșa and Transilvania Quantum’s Sorin Boloș join to discuss security vulnerabilities in quantum computing.
Investing in the invisible, ING plots a tech renaissance
Voice of the CTO: Less than a year in the job, Daniele Tonella delves into ING’s global data platform, gives his thoughts on the future of Agile development, and talks about the importance of “invisible controls” for tech development.
Evalueserve tames GenAI to boost client’s cyber underwriting
Firm’s insurance client adopts machine learning to interrogate risk posed by hackers