CTGT, a San Francisco-based startup specializing in AI for high-risk enterprise use cases, has unveiled an upgraded platform that eliminates bias, hallucinations, and other unwanted features from DeepSeek and other open-source AI models. Demonstrated at VentureBeat Transform 2025, the platform’s capabilities earned CTGT the top spot in the Innovation Showcase, highlighting its potential to transform AI reliability.
CTGT’s platform removes bias and hallucinations from DeepSeek and Llama models.
Achieves 96% response rate on sensitive queries, up from DeepSeek’s 32%.
Named top winner at VentureBeat Innovation Showcase 2025 in San Francisco.
Saved a leading insurer over $1M in financial exposure in one week.
Enables real-time compliance updates for banks without system downtime.
Backed by $7M from Gradient, Y Combinator, and notable AI investors.
AI trustworthiness remains a critical barrier for enterprises, with McKinsey reporting $67.4 billion in global losses due to AI hallucinations in 2024. Deloitte data indicates 47% of enterprise AI users have made decisions based on flawed outputs. “AI model trustworthiness is a major obstacle to enterprises achieving return on their AI investments,” said Cyril Gorlla, co-founder and CEO of CTGT. CTGT’s platform tackles these issues by identifying and neutralizing problematic model features, ensuring reliable outputs without costly retraining.
CTGT’s platform significantly enhances the performance of DeepSeek’s R1 model, which originally answered only 32% of sensitive queries due to a 14.9% hallucination rate and intentional bias coded into the model. “DeepSeek’s introduction of R1 drove many enterprises to reach out to us looking for help - they were interested in taking advantage of DeepSeek’s low cost and speed but concerned about bias,” said Gorlla. In testing, CTGT enabled DeepSeek to answer 96% of sensitive questions accurately, a marked improvement achieved in minutes without fine-tuning or backpropagation.
CTGT’s technology has delivered measurable results across industries. For a leading insurance provider, the platform ensured accurate coding of complex claims, avoiding over $1M in financial exposure in just one week. In the financial sector, CTGT supports a major bank by updating compliance models in real time, maintaining operational continuity. The platform’s efficiency stems from its ability to modify model behavior at inference time, reducing compute needs by an order of magnitude compared to traditional methods.
Launched in February 2025 with $7M in funding from investors like Gradient, General Catalyst, and Y Combinator, CTGT initially focused on reducing AI training costs. Its pivot to addressing bias and hallucinations has garnered attention from Fortune 500 executives. The platform’s showcase at VentureBeat Transform 2025, held June 24-25 in San Francisco, earned it the top Innovation Showcase award among seven finalists. CTGT’s approach, which also applies to models like Llama, positions it as a leader in making AI safer and more reliable for enterprise adoption.
CTGT’s platform represents a pivotal advancement in AI reliability, enabling enterprises to deploy trustworthy models in high-stakes environments. By eliminating bias and hallucinations in real time, CTGT is setting a new standard for AI safety and efficiency, driving confidence in enterprise AI investments.
CTGT enables enterprises to deploy AI for high-risk use cases, with a breakthrough mathematical method that removes hallucinations, censorship and bias at the model level automatically in real time, with no training or fine-tuning needed. Companies are using CTGT's platform to improve model accuracy, reduce compute and get models deployed up to 500x faster. Based on technology developed at the University of California at San Diego, CTGT is backed by General Catalyst, Y Combinator and notable angels including Paul Graham and François Chollet.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.