When Anthropic announced its AI model Mythos, the company was clear in saying that the model is too powerful for the world—and it’s only going to be available to select partners so that they can secure the systems that form the basis of the modern internet.
Much like Mythos, OpenAI is building an advanced AI system that we may soon see in testing. And right after that, Chinese players will likely be on the same road, maybe even open-sourcing the technology to make it available more widely.
While this has been the story of AI all these years, these releases will be a defining moment. We will soon have systems so powerful they could be exploited in novel ways to break global infrastructure, starting with the financial system we all depend on.
The world is not ready for that, but Charu Sharma is already racing to reinforce the banking system against the impending wave of AI-driven attacks.
As the CEO of Fenrock AI, a Y Combinator-backed (Winter 2026) deep-tech startup, Sharma is building the autonomous AI agents that will protect the global financial infrastructure from this next generation of sophisticated fraud.
Taking on this level of systemic risk in a market obsessed with consumer AI requires a very specific type of founder—one whose baseline for “fighting fires” is much more literal than the standard Silicon Valley metaphor. As a teenager, Sharma fought off train goons in Mumbai to save lives, an act of courage that earned her India’s National Bravery Award.
Today, that same baseline of courage fuels her approach to business.
“To me, building a startup is the epitome of courage,” Sharma told Future Nexus. “You’re imagining a market before it’s obvious to the world, making big non-consensus decisions that can make or break your startup, and day to day, you’re eating glass and powering through punches. You can’t do this without courage. It is a skill that I work hard on practicing every day.”
Why banks can’t wait
Having previously served as co-founder and president of the healthcare API startup Osana, and subsequently spending time on the other side of the table as an investor with the $3 billion VC fund SignalFire, Sharma developed a close eye for operating in regulated markets. But, as generative AI models experienced step-function advancements, she saw a massive gap in the financial sector: compliance and safety for banks.
“I’m quite concerned about the exponential rise in AI abuse and fraud that we’re about to experience. Fraudsters can pretend to be you so easily,” Sharma noted.
To combat this looming wave of fraud, she teamed up with a co-founder uniquely equipped for the challenge: Michael M., an AI veteran who built Apple’s very first privacy-preserving machine learning infrastructure. All the iPhones out there today benefit from his techniques, which allow machine learning models to be trained without ever exposing private user data.
Together, they worked to build an “AI workspace for banking back office,” which essentially deploys AI to help bank staffers combat the attacks driven by AI.
Currently, banks monitor customer transactions 24/7. When suspicious activity occurs, financial crime compliance analysts must manually investigate each alert and file reports with the government whenever there’s something fishy. The system is designed to stop activities like money laundering that fund drug cartels, terrorism, etc., but the sheer volume of data keeps human analysts entirely overwhelmed.
Fenrock’s AI workspace makes this investigation easier for the analysts. It deploys AI agents that look at each case, analyze it, and provide all the context and data points the analysts need to decide on the case—essentially helping them handle 10 to 20 times more alerts every single day.
Sharma clearly pointed out that these agents are not doing the entire job of these analysts, but only reducing their workload. “The AI agents get them 95-99% of the way there, and the human does a ‘review and verify.’ So, ultimately, the human analysts make the final call,” she said.
But automating with AI doesn’t mean cutting corners on compliance.
While accelerating analyst workflows and helping them secure the financial system is the primary focus of Fenrock’s AI, many may be worried about the risk of hallucinations or that the AI agents may provide incorrect information, leading to the mishandling of cases.
To ensure this doesn’t happen and federal agencies get a complete picture of the due diligence conducted on every loan or suspicious activity alert at the bank, Fenrock’s agents maintain a complete audit trail of their work, recording every single action they take.
“We’re designing our entire system around auditability,” Sharma explained. “Our agents automatically log all decisions, steps, and decision rationale. This is huge for banks. We’re actually helping them become more compliant with bulletproof proof of compliance.”
Boring by design
Even as most of the AI industry conversations revolve around flashy consumer AI models and products, Sharma is bullish on the prospects of her invisible, “boring” back-office technology making waves and having an actual impact in the long run.
In fact, she believes this is exactly where the next decacorn will be born.
“Regulated spaces will be the last ones to adopt AI since the barrier to entry is high and building for them requires deep domain expertise and building many integrations with industry-specific legacy systems. All of this requires a lot of hard work and even more patience. Most founders are not attracted to this, and if they do, they give up quickly. This means less competition and high defensibility business,” Sharma said, adding that she’s taking the long-term view to serve customers for 20+ years.
This customer-centric focus is another notable aspect of Fenrock’s approach. Instead of the usual playbook of performative networking, Sharma has built an obsession with her customers’ needs. She spends her time onsite, watching compliance officers copy and paste data hundreds of times just because “it’s part of the job” and obsessing over how to solve it for them.
“As a VC, I was more critical of ideas and tended to look for potential loopholes. This faded with time, especially as I intentionally spend every second I can with my customers (not other investors!) and am unreasonably close to their pain point,” Sharma said.
This proximity not only motivates Sharma to keep going but also breeds trust that can sometimes take an unexpected form. One past client even gifted Sharma her own wedding dress when the founder decided to elope at the last minute.
“I’m fascinated by banks. I am in love with the CEOs and compliance officers of small and mid-sized banks,” she said. “They are good-hearted people who have worked at their bank for 20-30 years, they really care about their customers, and they work really hard. I want to make them more successful so they can do what’s best for their customers and community.”
The weight of building
Building a generational company requires pacing. As a third-time founder, who presented at the W26 Demo Day, Sharma has learned to process the psychological toll of entrepreneurship.
She frequently recommends books like Wintering and Transitions to fellow founders—not for business strategy, but to power through fears and keep going.
She admits she tends to push too hard and burn out. Learning to view rest as joyful, rather than just a tool for better performance, has been vital. She has also learned to process the frequent “endings” in startup life, acknowledging the grief of losing free time and structure as a business scales.
Ultimately, the grueling daily grind of building Fenrock AI is anchored by the needs of everyday people. Financial crime compliance is about not only saving banks money but also protecting the livelihoods of everyday people from the very technology threats Anthropic and OpenAI are warning us about.
“We’re sort of in a race against time to secure our infrastructure to keep our money safe,” Sharma said. “My parents are getting older in India. I want their money to be safe, otherwise how will they pay for food and medical expenses? I run a business and am responsible for people’s livelihoods. This is really something that any human can relate to.”
This articles is written by : Fady Askharoun Samy Askharoun
All Rights Reserved to Amznusa www.amznusa.com
Why Amznusa?
AMZNUSA is a dynamic website that focuses on three primary categories: Technology, e-commerce and cryptocurrency news. It provides users with the latest updates and insights into online retail trends and the rapidly evolving world of digital currencies, helping visitors stay informed about both markets.
