When British doctor Ahmed Kerwan began working as a physician, the paperwork burden shocked him. On some days, he would spend only three hours actually caring for patients, with the rest of his workday spent on things like dealing with insurance claims.
There are already dozens, perhaps hundreds, of startups using AI to reduce the notoriously complex admin burden in healthcare. From note-taking specialists like Abridge to AI assistants startup Ambience, these startups are racing to streamline efficiencies. Kerwan, now an entrepreneur, is the founder and CEO of one of such company called Taxo. His startup offers an app that doctors and others use for tasks like getting prior authorizations from insurers, patient intake, and medical billing.
What sets Taxo apart, Kerwan says, is its AI “reasoning engine,” which transparently explains the process behind its decisions to users, helping it build trust with doctors. Reasoning models went mainstream in the AI world late last year by fleshing out their logic openly to users. At Taxo, the tech helps reduce hallucinations while increasing prior authorization approval rates to 98%, compared to an industry average of about 80%, according to Kerwan.
Taxo built its ‘reasoning engine’ by adding a healthcare-specific layer on top of existing models like OpenAI’s and Anthropic’s. It says the system is trained on hard-to-access healthcare data that makes it difficult for others to scrape overnight. “We didn’t want to be steamrolled every time OpenAI launches a new model,” Kerwan told TechCrunch.
The reasoning trend in AI remains early and only really gained traction with the rise of Chinese startup DeepSeek. Investor interest in Taxo, however, suggests there’s a chance for the technique to gain broader adoption beyond foundational AI companies. The startup recently closed a $5 million seed round led by Y Combinator, General Catalyst, and Character Capital.
Founded last year and based in San Francisco, Taxo tells TechCrunch it passed $1 million ARR six months after its launch. It now serves about 15 customers, ranging from clinics to government providers.
When ChatGPT was released, doctors were understandably cautious about using it because they couldn’t trace why and how it was making specific recommendations, Kerwan told TechCrunch. He’s hoping Taxo changes that. “You can see exactly where we got the information and why it’s being given,” he said.
Keep reading the article on Tech Crunch
The encrypted messaging app Signal is getting some unexpected attention this week.
High-ranking officials in the Trump administration, including Vice President J. D. Vance and Secretary of Defense Peter Hegseth, communicated the plans for an attack on the Yemeni Houthis via a potentially unauthorized group chat on Signal. However, Atlantic editor-in-chief Jeffrey Goldberg was mistakenly added to the group chat, giving him access to these highly sensitive discussions, which he later published.
The Signal app itself did not malfunction or operate in an unintended way. Rather, it is user error to accidentally add a journalist to a chat about U.S. military plans — an error that government security protocols should be able to prevent if they’re actually followed.
When the Atlantic’s story broke on Monday, worldwide Signal downloads on iOS and Google Play were up 28% from the daily average over the last 30 days, per app intelligence firm Appfigures. In the U.S., downloads were up 45% on Monday, and in Yemen, they were up by 42%. Before the scandal, Signal was ranked No. 50 among social media apps in Yemen, but it climbed to No. 9 on Monday.
Signal did not respond to TechCrunch’s request for comment.
All communications on Signal are encrypted, meaning that only the people in a chat can see the texts — not even people who work at Signal can know what users are talking about. But Signal is intended to be a consumer product for secure messaging, not an iron-clad depository for government military plans.
Although Hegseth said that there were “no war plans” discussed in the Signal chat, the Atlantic published messages that show Hegseth providing details about the timing of attacks, as well as the weapons and aircrafts that would be used.
As of Thursday, the government continues to investigate this monumental security failure.
Keep reading the article on Tech Crunch
Mexico City-based Mendel has raised $35 million in a Series B round of funding, it tells TechCrunch exclusively.
Corporate spend management platform Mendel last raised in December 2021 — a $15 million Series A round and $20 million in debt — after participating in Y Combinator’s Winter 2021 cohort. With this latest capital infusion, the startup has brought in a total of $60 million in equity funding and $50 million via a credit facility.
Mendel’s mission is straightforward: to reinvent corporate spend management by automating most of the operations for an enterprise CFO that are currently done manually. Or put even more simply, it wants to be a one-stop shop for all B2B spend. Its offering integrates expense management, payments, and corporate travel.
“Our goal is to give CFOs and finance teams in Latin America real-time visibility and control over their spend — be it employee expenses, vendor payments, or business travel bookings,” said co-CEO and co-founder Alan Karpovsky.
Karpovsky and Alejandro Zecler (who both previously founded and sold other startups) started Mendel in early 2021, and Helena Polyblank (CPO) and Gonzalo Castiglione (CTO) later joined as co-founders.
Mendel declined to reveal valuation, with Karpovsky saying only the round reflected “a significant step up” from the company’s previous raise. The company also declined to reveal hard revenue figures, with Karpovsky noting only that its annual recurring revenue (ARR) grew almost 2.5x year-over-year, with gross margins of over 75%.
“We’re not yet profitable, but we anticipate reaching profitability by late 2025,” he told TechCrunch.
Base10 Partners led Mendel’s latest round, which included participation from new investors PayPal Ventures and Endeavor Catalyst, as well as existing backers Infinity Ventures, Industry Ventures, and Hi.vc.
The company says that since it is “software first” and focused on enterprises, it is able to charge recurring SaaS fees rather than relying exclusively on interchange revenue or lending-based models. Its revenue comes from a combination of SaaS fees (over 50%) for its expense management and travel tool and interchange fees from credit cards as well as a take rate from its bill pay product.
Karpovsky believes that the company’s LatAm focus gives it an advantage over other global players in that it’s able to address “complex, country-specific regulations” such as tax codes, invoicing requirements, and multi-currency workflows, among other things.
“We like to say ‘Mendel is like SAP Concur and AMEX having a child,’ Karpovsky quipped.
As for comparisons to New York-based decacorn Ramp, he said that in many ways, “Mendel is like Ramp for Latin American enterprises” with a few differentiators, including the fact that it is focused on “large, complex organizations that require multi-entity, multi-currency, multi-credit-line and deep ERP integrations.”
Presently, Mendel has 80 employees, compared to 64 employees a year ago. Looking ahead, the company plans to expand geographically. It’s already operating in Mexico and Argentina with about 500 customers, including Mercado Libre, FEMSA, Adecco, and McDonald’s. It’s looking to expand into Chile, Colombia, and Peru in 2025, and Brazil in 2026.
“Our approach from day zero was first consolidating the largest Spanish-speaking market in LatAm before starting the geo-expansion,” Karpovsky said.
Base10 Partner Jason Kong told TechCrunch that his firm was attracted to what it viewed as Mendel’s “unique positioning” as a spend management platform for large companies in underserved — but growing — Latin America.
“The company’s high capital efficiency — being cash-flow positive in December 2024 — stood out in a sector where many players struggle with unit economics,” Kong added. “Additionally, Mendel’s ability to replace legacy solutions like SAP Concur and win large enterprise customers at a fast sales velocity (sub-3 months for 3,000+ employee enterprises) demonstrated clear product-market fit.” Other companies also operating in this space in Latin America include Clara and Jeeves — another YC alum — but both target more SMBs and rely more on transactional fees, noted Kong.
Keep reading the article on Tech Crunch
Abound, a remittance app that was spun off by Times Internet in 2023, has raised $14 million in its first external funding round as it aims to reach more Indian expats in the U.S.
Remittance flows to India are rising as the Indian diaspora spreads worldwide. In 2024, the South Asian country recorded $129.1 billion in remittances, accounting for 14.3% share of the global market and topping the charts, according to a World Bank report. Abound aims to tap this growth with its mobile app.
“Indians are among the largest immigrant groups in the U.S. The average household income in the U.S. is close to $58,000, and the average Indian household income is about $150,000. That tells you that Indian expats are wealthy, affluent, and yet they’re vastly underserved in terms of products and services that are geared for them,” said Nishkaam Mehta, CEO of Abound, in an interview.
Mehta, who worked at Hulu as head of its mobile strategy and growth for more than four years, joined Times Internet in 2019 after meeting its vice chairman Satyan Gajwani to create a “super app” for non-resident Indians. The startup was incubated at the tech arm of Indian media conglomerate, The Times of India Group.
Initially named Times Club, Abound allows users to send money to India, earn rewards, and get cashback on services including live sports streaming, grocery shopping, and OTT subscriptions. The firm has plans to explore avenues to let users access high-yield savings, India-focused investments, and cross-border credit solutions.
“In our model as a super app, we envision a role for banks themselves to be a part of the platform,” Mehta told TechCrunch.
The company claims it has processed over $150 million in remittances in total from its more than 500,000 monthly transacting users, and that its revenue has increased by 50% month-over-month since launch.
Abound’s remittance volume increased by 15% every month and the startup processed $110 million to $120 million in the past 12 months, Mehta said.
Abound generates ad revenue from rewards and foreign exchange spread on money remittances. Foreign exchange presents significant potential for growth, Mehta stated. The startup said The Times of India’s over 50 million monthly online visitors outside India also help it reach new users and offer a range of rewards.
“In money remittances, if you purely play the exchange rate game, then you’re always acquiring the user,” said Mehta. “In our case, because we’ve got this rewards layer from the Times of India and other local advertisers, we don’t have that problem. We can always compete on exchange rates, knowing that we don’t have the same customer acquisition cost that the other companies might have.”
This seed round was all-equity, and was led by NEAR Foundation, with participation from Circle Ventures, Times Internet, and other investors. The company plans to use the fresh cash to expand its presence, increase its offerings and improve its tech infrastructure.
“Traditional banks in the U.S. don’t focus on the financial requirements of this segment because there is no banking product built just for the NRI population. We see that as a large gap and opportunity,” said Gajwani.
Following the deal, Times Internet will continue to be the largest stakeholder in Abound. Gajwani told TechCrunch the Times Internet would be “using its strategic assets to help accelerate Abound’s growth.”
The market of platforms enabling foreign remittances is crowded with incumbents such as Western Union, PayPal and MoneyGram, as well as newer players like Remitly and Wise. But Mehta thinks Abound has an edge as it “super serves” users by offering competitive exchange rates as well as rewards and cashback at about 5,000 Indian grocery stores and access to live-streamed cricket — by far the most popular sport in India.
Abound currently has a team of 40 people, primarily based in India. It plans to expand its headcount and set up an executive team in the U.S. as well.
In time, the firm plans to enter markets such as Canada, Singapore and the UAE, which all have big populations of non-resident Indians. Nonetheless, Mehta said the immediate focus is to cement its footing in the U.S. and then run pilots in foreign markets.
Keep reading the article on Tech Crunch
UnitedHealth Group has scrubbed much of its website mentioning its diversity, equity, and inclusion (DEI) policies, including pulling down blog posts and removing large sections from its website, TechCrunch has learned.
According to archived copies of UnitedHealth’s website, several of the company’s web pages dedicated to DEI no longer load and now redirect to a “page not found” error. A section of the company’s career page that used to have a dedicated section for diversity, equity, and inclusion, along with its diversity initiatives, no longer appears on the same live page. UnitedHealth also removed a 2022 blog post featuring a conversation with its vice president of DEI.
It’s not clear why UnitedHealth pulled down the pages, and if it represents a shift of verbiage or an actual change in its policies. UnitedHealth spokesperson Tyler Mason did not return requests for comment Wednesday
The removal of DEI from UnitedHealth’s websites coincides with a broad retreat from DEI policies and programs by household names and tech companies alike, amid mounting pressure from Trump administration-issued executive orders targeting DEI programs.
In February, U.S. Attorney General Pam Bondi instructed the Justice Department to “investigate, eliminate, and penalize” DEI programs that it considers illegal at private sector companies that receive federal funding. A federal appeals court temporarily allowed the Trump administration to press ahead, despite a lower court ruling the government’s efforts unlawful.
Several tech companies have already scrubbed mentions of DEI from their websites, including Google and OpenAI.
TechCrunch saw UnitedHealth take down its web pages mentioning DEI throughout Wednesday morning in real time. Since the ransomware attack on Change Healthcare last year, TechCrunch has used a web page monitor to automatically and continually check for any changes to its website, such as updates to its data breach notice.
In removing “Diversity, Equity and Inclusion” in UnitedHealth’s website menu, the company added a slimmed-down page with weaker language it calls the “Culture of Belonging,” which leaves out previous references to the company’s diversity efforts on university campuses, diversity in recruiting, and the company’s various employee resource groups.
Keep reading the article on Tech Crunch
A complaint about poverty in rural China. A news report about a corrupt Communist Party member. A cry for help about corrupt cops shaking down entrepreneurs.
These are just a few of the 133,000 examples fed into a sophisticated large language model that’s designed to automatically flag any piece of content considered sensitive by the Chinese government.
A leaked database seen by TechCrunch reveals China has developed an AI system that supercharges its already formidable censorship machine, extending far beyond traditional taboos like the Tiananmen Square massacre.
The system appears primarily geared toward censoring Chinese citizens online but could be used for other purposes, like improving Chinese AI models’ already extensive censorship.
Xiao Qiang, a researcher at UC Berkeley who studies Chinese censorship and who also examined the dataset, told TechCrunch that it was “clear evidence” that the Chinese government or its affiliates want to use LLMs to improve repression.
“Unlike traditional censorship mechanisms, which rely on human labor for keyword-based filtering and manual review, an LLM trained on such instructions would significantly improve the efficiency and granularity of state-led information control,” Qiang told TechCrunch.
This adds to growing evidence that authoritarian regimes are quickly adopting the latest AI tech. In February, for example, OpenAI said it caught multiple Chinese entities using LLMs to track anti-government posts and smear Chinese dissidents.
The Chinese Embassy in Washington, D.C., told TechCrunch in a statement that it opposes “groundless attacks and slanders against China” and that China attaches great importance to developing ethical AI.
The dataset was discovered by security researcher NetAskari, who shared a sample with TechCrunch after finding it stored in an unsecured Elasticsearch database hosted on a Baidu server.
This doesn’t indicate any involvement from either company — all kinds of organizations store their data with these providers.
There’s no indication of who, exactly, built the dataset, but records show that the data is recent, with its latest entries dating from December 2024.
In language eerily reminiscent of how people prompt ChatGPT, the system’s creator tasks an unnamed LLM to figure out if a piece of content has anything to do with sensitive topics related to politics, social life, and the military. Such content is deemed “highest priority” and needs to be immediately flagged.
Top-priority topics include pollution and food safety scandals, financial fraud, and labor disputes, which are hot-button issues in China that sometimes lead to public protests — for example, the Shifang anti-pollution protests of 2012.
Any form of “political satire” is explicitly targeted. For example, if someone uses historical analogies to make a point about “current political figures,” that must be flagged instantly, and so must anything related to “Taiwan politics.” Military matters are extensively targeted, including reports of military movements, exercises, and weaponry.
A snippet of the dataset can be seen below. The code inside it references prompt tokens and LLMs, confirming the system uses an AI model to do its bidding:
From this huge collection of 133,000 examples that the LLM must evaluate for censorship, TechCrunch gathered 10 representative pieces of content.
Topics likely to stir up social unrest are a recurring theme. One snippet, for example, is a post by a business owner complaining about corrupt local police officers shaking down entrepreneurs, a rising issue in China as its economy struggles.
Another piece of content laments rural poverty in China, describing run-down towns that only have elderly people and children left in them. There’s also a news report about the Chinese Communist Party (CCP) expelling a local official for severe corruption and believing in “superstitions” instead of Marxism.
There’s extensive material related to Taiwan and military matters, such as commentary about Taiwan’s military capabilities and details about a new Chinese jet fighter. The Chinese word for Taiwan (台湾) alone is mentioned over 15,000 times in the data, a search by TechCrunch shows.
Subtle dissent appears to be targeted, too. One snippet included in the database is an anecdote about the fleeting nature of power which uses the popular Chinese idiom, “when the tree falls, the monkeys scatter.”
Power transitions are an especially touchy topic in China thanks to its authoritarian political system.
The dataset doesn’t include any information about its creators. But it does say that it’s intended for “public opinion work,” which offers a strong clue that it’s meant to serve Chinese government goals, one expert told TechCrunch.
Michael Caster, the Asia program manager of rights organization Article 19, explained that “public opinion work” is overseen by a powerful Chinese government regulator, the Cyberspace Administration of China (CAC), and typically refers to censorship and propaganda efforts.
The end goal is ensuring Chinese government narratives are protected online, while any alternative views are purged. Chinese President Xi Jinping has himself described the internet as the “frontline” of the CCP’s “public opinion work.”
The dataset examined by TechCrunch is the latest evidence that authoritarian governments are seeking to leverage AI for repressive purposes.
OpenAI released a report last month revealing that an unidentified actor, likely operating from China, used generative AI to monitor social media conversations — particularly those advocating for human rights protests against China — and forward them to the Chinese government.
If you know more about how AI is used in state opporession, you can contact Charles Rollet securely on Signal at charlesrollet.12 You also can contact TechCrunch via SecureDrop.
OpenAI also found the technology being used to generate comments highly critical of a prominent Chinese dissident, Cai Xia.
Traditionally, China’s censorship methods rely on more basic algorithms that automatically block content mentioning blacklisted terms, like “Tiananmen massacre” or “Xi Jinping,” as many users experienced using DeepSeek for the first time.
But newer AI tech, like LLMs, can make censorship more efficient by finding even subtle criticism at a vast scale. Some AI systems can also keep improving as they gobble up more and more data.
“I think it’s crucial to highlight how AI-driven censorship is evolving, making state control over public discourse even more sophisticated, especially at a time when Chinese AI models such as DeepSeek are making headwaves,” Xiao, the Berkeley researcher, told TechCrunch.
Keep reading the article on Tech Crunch