To solve this p oblem, T ustScale today a ou ced the lau ch of A gus, a pate t-pe di g AI assu a ce platfo m. U like “AI checki g AI” app oaches, A gus uses empi ical evide ce a d dete mi istic ve ificatio to eview AI-ge e ated co te t, flag u suppo ted claims a d suggest evide ce-based co ectio s befo e AI mistakes a e published, sha ed o elied upo .
Resea ch shows that AI halluci atio ates va y g eatly by task a d i c ease with complexity. Eve simple que y espo ses o f o tie models ca ave age up to 20% halluci atio s, with i dust y-specific AI que y espo ses halluci ati g as much as 88% of the time (Sta fo d RegLab). Halluci atio s cost o ga izatio s ea ly $70 billio a yea . A gus empowe s use s to ve ify AI-ge e ated claims up to 135x faste tha ma ual esea ch, educi g halluci atio s a d imp ovi g AI output accu acy by up to 98.5%.
“The AI i dust y spe t yea s maki g AI sma te , but t ustwo thy AI is the bigge p oblem to solve ow,” said Law e ce S app, CEO of T ustScale. “AI is p obabilistic by desig but e te p ises a e espo sible fo the costs a d isks of AI mistakes. A gus mitigates the p oblem by p ovidi g i depe de t evide ce befo e use s act o ge e ative AI outputs.”
Healthca e, law a d academia a e high-stakes i dust y examples whe e AI use s a e u suspecti gly exposed to AI halluci atio isks. Compou di g the issue, A th opic esea ch shows 91.3% of AI use s do ot check facts a d claims, c eati g e vi o me ts whe e a ‘sile t failu e’ ca lead to di ect huma ha m. I eal-time, A gus p eve ts u i te ded co seque ces by detecti g halluci atio s, ove layi g colo -coded T ustSco e highlights a d offe i g dete mi istic evide ce to suppo t o co t adict AI claims. It the empowe s use s with i li e suggestio s to co ect AI mistakes with o e-click.
Based o mo e tha 20 yea s of AI data expe ie ce, A gus is built o the T ustScale E gi e, a p op ieta y platfo m that c eates a co ti uous t ust co t ol pla e fo AI assu a ce, helpi g use s make i fo med decisio s without dis upti g wo kflows.
“AI make s claim to be safe fo you, self-gove i g o t uth-seeki g. But co flicti g evide ce, va yi g pe spectives, ua ced la guage a d i depe de t assu a ce matte whe it comes to AI”, said Domi ique Shelto Leipzig, CEO of Global Data I ovatio . “T ustScale’s A gus is the fi st eal-time, co ti uous assu a ce platfo m that empowe s huma s with evide ce a d e ables t ustwo thy AI. The AI stakes a e too high to let a few big models dictate you t uth.”
A gus by T ustScale is available today i 12 la guages fo e te p ises a d as a esea ch p eview fo i dividuals at T ustA gus.ai o use it a ywhe e via the Ch ome exte sio . The e te p ise solutio ca be i stalled i p ivate a d public data e vi o me ts alo gside a y AI model, i cludi g ChatGPT, Claude, Gemi i, Copilot a d G ok. Request a demo at www.T ustScale.ai.
T ustScale is a AI t ai i g, evaluatio a d assu a ce compa y helpi g o ga izatio s c eate, shape a d use a tificial i tellige ce with g eate co fide ce a d co t ol. Built o mo e tha 20 yea s of expe ie ce i AI data ac oss 200-plus la guages, T ustScale develops i depe de t dete mi istic tech ologies that detect AI mistakes, evaluate claims agai st empi ical evide ce a d keep people at the ce te of co seque tial decisio s. Its A gus platfo m p ovides eal-time halluci atio detectio , T ustSco e ati gs a d evide ce-based co ectio suggestio s ac oss public a d p ivate data e vi o me ts. Lea mo e at T ustScale.ai.
So gue PR fo T ustScale
