TechCrunch+

California’s Department of Finance has confirmed it’s investigating a “cybersecurity incident” after the prolific LockBit ransomware group claims to have stolen confidential data from the agency.

The California Office of Emergency Services (Cal OES) in a statement on Monday described the threat as an “intrusion” that was “identified through coordination with state and federal security partners.”

The statement did not provide any specifics about the nature of the incident, who was involved, or whether any information had been stolen. The California Department of Finance did not respond to TechCrunch’s questions prior to publication.

“While we cannot comment on specifics of the ongoing investigation, we can share that no state funds have been compromised, and the department of finance is continuing its work to prepare the governor’s budget that will be released next month,” the statement said.

While state officials remain tight-lipped about the incident, the notorious LockBit ransomware gang on Monday claimed responsibility for the attack. In a post on its dark web leak site seen by TechCrunch, the Russia-affiliated group claims to have stolen 76GB of files from the agency, including “databases, confidential data, financial documents, certification, IT documents, and sexual proceedings in court.”

Screenshots shared by LockBit lend some weight to its claim, but the ransomware gang’s claims should still be taken with skepticism. In June, the group claimed it breached cybersecurity company Mandiant, which was later revealed as false. The ransomware group faked the incident in response to a Mandiant investigation that demonstrated significant overlaps between LockBit and the U.S.-sanctioned Evil Corp group.

LockBit has given California’s finance department a December 24 deadline to pay its as-yet unspecified ransom demand. If the agency fails to pay, the ransomware gang is threatening to leak the entire cache of stolen data.

This latest breach comes just weeks after the U.S. Department of Justice in November charged a dual Russian and Canadian citizen linked to LockBit over his alleged involvement in attacks targeting critical infrastructure and large industrial groups worldwide. At the time, the DOJ said that LockBit has claimed at least 1,000 victims in the United States and has extracted tens of millions of dollars in actual ransom payments from their victims.

California’s finance department confirms breach as LockBit claims data theft by Carly Page originally published on TechCrunch

Image-generating AI models like DALL-E 2 and Stable Diffusion can — and do — replicate aspects of images from their training data, researchers show in a new study, raising concerns as these services enter wide commercial use.

Co-authored by scientists at the University of Maryland and New York University, the research identifies cases where image-generating models, including Stable Diffusion, “copy” from the public internet data — including copyrighted images — on which they were trained.

The study hasn’t been peer reviewed yet, and the co-authors submitted it to a conference whose rules forbid media interviews until the research has been accepted for publication. But one of the researchers, who asked not to be identified by name, shared high-level thoughts with TechCrunch via email.

“Even though diffusion models such as Stable Diffusion produce beautiful images, and often ones that appear highly original and custom tailored to a particular text prompt, we show that these images may actually be copied from their training data, either wholesale or by copying only parts of training images,” the researcher said. “Companies generating data with diffusion models may need to reconsider wherever intellectual property laws are concerned. It is virtually impossible to verify that any particular image generated by Stable Diffusion is novel and not stolen from the training set.”

Images from noise

State-of-the-art image-generating systems like Stable Diffusion are what’s known as “diffusion” models. Diffusion models learn to create images from text prompts (e.g., “a sketch of a bird perched on a windowsill”) as they work their way through massive training data sets. The models — trained to “re-create” images as opposed to drawing them from scratch — start with pure noise and refine an image over time to make it incrementally closer to the text prompt.

It’s not very intuitive tech. But it’s exceptionally good at generating artwork in virtually any style, including photorealistic art. Indeed, diffusion has enabled a host of attention-grabbing applications, from synthetic avatars in Lensa to art tools in Canva. DeviantArt recently released a Stable Diffusion–powered app for creating custom artwork, while Microsoft is tapping DALL-E 2 to power a generative art feature coming to Microsoft Edge.

To be clear, it wasn’t a mystery that diffusion models replicate elements of training images, which are usually scraped indiscriminately from the web. Character designers like Hollie Mengert and Greg Rutkowski, whose classical painting styles and fantasy landscapes have become one of the most commonly used prompts in Stable Diffusion, havedecried what they see as poor AI imitations that are nevertheless tied to their names.

But it’s been difficult to empirically measure how often copying occurs, given diffusion systems are trained on upward of billions of images that come from a range of different sources.

To study Stable Diffusion, the researchers’ approach was to randomly sample 9,000 images from a data set called LAION-Aesthetics — one of the image sets used to train Stable Diffusion — and the images’ corresponding captions. LAION-Aesthetics contains images paired with text captions, including images of copyrighted characters (e.g., Luke Skywalker and Batman), images from IP-protected sources such as iStock, and art from living artists such as Phil Koch and Steve Henderson.

The researchers fed the captions to Stable Diffusion to have the system create new images. They then wrote new captions for each, attempting to have Stable Diffusion replicate the synthetic images. After comparing using an automated similarity-spotting tool, the two sets of generated images — the set created from the LAION-Aesthetics captions and the set from the researchers’ prompts — the researchers say they found a “significant amount of copying” by Stable Diffusion across the results, including backgrounds and objects recycled from the training set.

One prompt — “Canvas Wall Art Print” — consistently yielded images showing a particular sofa, a comparatively mundane example of the way diffusion models associate semantic concepts with images. Others containing the words “painting” and “wave” generated images with waves resembling those in the painting “The Great Wave off Kanagawa” by Katsushika Hokusai.

Across all their experiments, Stable Diffusion “copied” from the training data set roughly 1.88% of the time, the researchers say. That might not sound like much, but considering the reach of diffusion systems today — Stable Diffusion had created over 170 million images as of October, according to one ballpark estimate — it’s tough to ignore.

“Artists and content creators should absolutely be alarmed that others may be profiting off their content without consent,” the researcher said.

Implications

In the study, the co-authors note that none of the Stable Diffusion generations matched their respective LAION-Aesthetics source image and that not all models they tested were equally prone to copying. How often a model copied depended on several factors, including the size of the training data set; smaller sets tended to lead to more copying than larger sets.

One system the researchers probed, a diffusion model trained on the open source ImageNet data set, showed “no significant copying in any of the generations,” they wrote.

The co-authors also advised against excessive extrapolation from the study’s findings. Constrained by the cost of compute, they were only able to sample a small portion of Stable Diffusion’s full training set in their experiments.

Still, they say that the results should prompt companies to reconsider the process of assembling data sets and training models on them. Vendors behind systems such as Stable Diffusion have long claimed that fair use — the doctrine in U.S. law that permits the use of copyrighted material without first having to obtain permission from the rightsholder — protects them in the event that their models were trained on licensed content. But it’s an untested theory.

“Right now, the data is curated blindly, and the data sets are so large that human screening is infeasible,” the researcher said. “Diffusion models are amazing and powerful, and have showcased such impressive results that we cannot jettison them, but we should think about how to keep their performance without compromising privacy.”

For the businesses using diffusion models to power their apps and services, the research might give pause. In a previous interview with TechCrunch, Bradley J. Hulbert, a founding partner at law firm MBHB and an expert in IP law, said he believes that it’s unlikely a judge will see the copies of copyrighted works in AI-generated art as fair use — at least in the case of commercial systems like DALL-E 2. Getty Images, motivated out of those same concerns, has banned AI-generated artwork from its platform.

The issue will soon play out in the courts. In November, a software developer filed a class action lawsuit against Microsoft, its subsidiary GitHub and business partner OpenAI for allegedly violating copyright law with Copilot, GitHub’s AI-powered, code-generating service. The suit hinges on the fact that Copilot — which was trained on millions of examples of code from the internet — regurgitates sections of licensed code without providing credit.

Beyond the legal ramifications, there’s reason to fear that prompts could reveal, either directly or indirectly, some of the more sensitive data embedded in the image training data sets. As a recent Ars Technica report revealed, private medical records — as many as thousands — are among the photos hidden within Stable Diffusion’s set.

The co-authors propose a solution in the form of a technique called differentially private training, which would “desensitize” diffusion models to the data used to train them — preserving the privacy of the original data in the process. Differentially private training usually harms performance, but that might be the price to pay to protect privacy and intellectual property moving forward if other methods fail, the researchers say.

“Once the model has memorized data, it’s very difficult to verify that a generated image is original,” the researcher said. “I think content creators are becoming aware of this risk.”

Image-generating AI can copy and paste from training data, raising IP concerns by Kyle Wiggers originally published on TechCrunch

Elon Musk-led Twitter is shaking up its verification system. Instead of one checkmark, now there are multicolored checkmarks to denote different things. This could be very confusing for users to track. So here’s a handy guide to all checkmarks and badges on the social network.

Checkmarks

Blue checkmark: It currently means two things. 1) Account with this checkmark is a legacy verified account (read: verified in the pre-Musk era). This was used to mark a notable account representing a politician, a celebrity, or an activist. This was to prove that the said person is indeed who they are claiming to be. Musk has said that the legacy checkmark will go away in a few months. 2) Account with the blue check mark can also mean that the person has subscribed to it. The only way to know the difference between the two of them is to click on the blue checkmark.

Gold checkmark: Twitter debuted this checkmark earlier this week to note that the account belongs to a company or an organization. The social network also said that it is working on a Twitter Blue for the Business plan so companies can apply to get a checkmark.

Official (Grey) checkmark: This newly introduced secondary checkmark is a way to certify certain profiles, such as accounts from governments, political parties, media houses, and brands. Twitter also says this applies to “some other public figures” without any specification.

The official checkmark seemingly serves the same purpose as the legacy verification system. But it can exist alongside a blue or a gold checkmark.

Labels and badges

State-affiliated media:Twitter applies this label (podium icon) to media houses that don’t have editorial independence. That means the state has editorial and financial control of that media entity. So, entities like BBC and NPR don’t come under that purview. The label, introduced two years ago, applies to media accounts along with their prominent editors and reports. The social network also doesn’t amplify or recommend these accounts or their tweets.

Government accounts:Introduced along with state-affiliated media labels, the government account label (flag icon) aims to notify that this is an account belonging to a government entity or operated by an official. Twitter doesn’t label accounts that are not official communication channels for government personnel.

Twitter says that in limited cases, where a government is depressing people’s voices, it doesn’t recommend of amplifying that account.

US election candidates:This label notes that the account belongs to a person who is participating in the US midterms or running for House of Representatives, U.S. Senate, or Governor. The company has used these labels in multiple elections now.

Automated labels:Twitter started testing a label for good and useful bots last year and officially launched it this February. However, given Musk’s “war against bots,” there is no guarantee that this label will survive.

Verified phone number badge: Twitter had been testing this label before Musk took over. But recently, the company started rolling it out for users in India. Technically, Twitter doesn’t assign this label to anyone. If a user in India has verified their phone number, they can choose to show it on their profile.

Twitter profile category: This is another self-attested label for businesses. That means professional accounts can identify themselves with labels like “Coffee Shop,” “Journalist” and “Optician” (see the example above).

As Twitter is a Musk-led company, we never know when we will see a new label or the removal of one of the existing ones. We’ll keep this story updated for that situation.

A quick guide to all the checkmarks and badges on Twitter by Ivan Mehta originally published on TechCrunch

Artificial intelligence is touching every aspect of how we engage with information (and much more) these days. Today, a startup building out a business based on one particular application of that — how to apply AI to knowledge management in the workplace — is announcing some funding as it finds some decent traction for its approach. Sana Labs — which provides an AI-based platform to help people manage information at work, and subsequently to use that data as a resource for e-learning within the organization — has closed a round of $34 million after seeing ARR grow seven-fold in the last year.

Menlo Ventures, the U.S. VC firm, is leading the round for Stockholm-based Sana, with EQT Ventures and a whopping 25 angels and founder/operator individuals also participating. This is a Series B that values Sana at $180 million post-money.

There are a lot of knowledge management, enterprise learning and enterprise search products on the market today, but what Sana believes it has struck on uniquely is a platform that combines all three to work together: a knowledge management-meets-enterprise-search-meets-e-learning platform.

The crux of Sana is a platform and AI engine that connects to all of the different apps that an organization uses in the workplace — Salesforce, e-mail, Notion, Github, Slack, Trello, Asana, and whatever else you might have to capture, source or store information and communicate with others.

All of the data across these apps is ingested and organized automatically by the Sana platform (AI magic), and maintained as the information inside those apps changes or expands. Then, users who want to access information go to Sana and request it in regular “human” language as you might do in a search engine. But alongside that, the data is used to as the basis of e-learning modules for onboarding, training or professional development — modules created/conceived of either by people in the organization, or by Sana itself.

This wasn’t the original concept for Sana, which started with building just the back-end machine learning engine to organize information. But Joel Hellermark, Sana’s CEO and founder, said that early on the startup was getting requests for the front end — the part for people to easily query the information and use it to build training and learning materials — so they build that part, too. The learning can come in the form of quizzes and polls, interactive sessions, and more, and when interactive Q&A is generated around webinars, like some kind of very resourceful, waste-not-want-not stew, the outcomes from all those also get fed into the knowledge base for future reference.

The mix of knowledge management with search and e-learning means that the platform sees very different engagement metrics, said Hellermark. “Sana is used continuously which is very different from a typical e-learning platform,” he said. “We’re seeing weekly and daily active usage” from among the tens of thousands of employees from across the 100 or so businesses that are already using Sana, he added.

The tech itself is built out and customized by Sana, but the models, Hellermark said, come from OpenAI, which has a “deep partnership” with Sana, in Hellermark’s words.

“We’ve been using their models from day one continuously, since before launch,” he said. That includes GPT, which — via ChatGPT — has been the talk of the town among tech and media folk on chatty platforms like Twitter. Sana’s approach speaks to the scalable potential for AI longer term.

“We believe there will be underlying models from the likes of OpenAI with the opportunity to fine-tune them for specific domains,” Hellermark added. “For us, the focus is the user experience on top of this.”

Hellermark describes himself as a longtime obsessive about not just the significance of education, but of the power of AI to make a mark in the space. But education comes in many forms — content aimed at younger people, further education, adult learning, and professional development being just a few of the slices of the pie.

He said that Sana chose to focus on the fourth of those for two reasons. The first is because of the practicality of it — there isn’t really anything else like it on the market today, but it’s definitely something organizations could use, given the oversupply of useful information contained within an organization’s braintrust that works on an inverse variation: the more of it that is amassed, the harder it gets to tap into it.

The second reason for the enterprise focus is because of the scalability factor: while education in the more traditional sense clearly could use tools for ingesting lots of disparate, fragmented information and making it easily accessible and the basis of learning modules personalized to the individual, the fragmentation across age groups and school districts, let alone countries and their own specific curriculums, makes it a more complicated target — perhaps even more right now, given the emphasis we’re seeing from startups, and their backers, to focus on projects with sound unit economics, identifiable (and active) customer bases, and tech that already works to those ends.

“The education sector is my biggest passion because if you solve learning you solve everything,” he said. “But from day one we wanted to be a large company and it’s hard to scale that in K-12 because you have to adapt to different countries. Having an enterprise approach helps us scale and helps doctors to engineers and product managers and sales reps and everyone. We’re able to serve all of them in over 20 countries.”

Importantly, that’s not to say that this won’t be a target longer term, or that the traditional sector of education wouldn’t or couldn’t be a receptive customer for technology like this — from Sana or another startup — in the longer term.

Another important detail to consider is how Sana handles the quality of the information that it sources. How does it decide — can it decide? — if data that it sources is correct, and what does it do if there are multiple “answers” that are not consistent with each other?

“That is what knowledge management is,” Hellermark said in response to the question. “You can have models that are just search, but that doesn’t take into account the need to verify knowledge and create journeys.” He said that a “structure for verification” is built into the system, which includes people being able to limit what sources and other input can be used by Sana, with customers able choose to designate what information is verified and accurate, and choose whether users can access information that is unverified, and to rank information.

It’s not a fully satisfactory answer, to be honest, especially since accuracy one of the most persistent issues around AI: what do you do if it’s not quite right, or outright wrong, or simply using bad data?

As with the rest of the rocket ship that is AI, however, this for now has not been an issue impeding Sana’s growth.

“Over the past 6+ years, I’ve looked at almost every single other learning management system SaaS, and the best part about Sana is that they are building a true knowledge management solution from the ground up, considering how knowledge is captured in today’s knowledge economy,” said JP Sanday, the Menlo partner who led this investment. “Companies are now more distributed, are being asked to do more with less and cannot keep up with the pace of innovation and need to enable all of their employees. Sana is the only platform I have ever seen that can fulfill this vision.”

He added that the approach of people both tapping into the database, and building content around it, creates a specific “organizational knowledge graph” that is more democratized than what you typically get in organizations.

“When I show prospects the product and they see the content creation experience as well as the AI capabilities that help both authors and learners they immediately know they are looking at something completely different — they see how much more extensible it is and how much more engagement they get from users,” he said.

Sana raises $34M for its AI-based knowledge management and learning platform for workplaces by Ingrid Lunden originally published on TechCrunch

It doesn’t matter what kind of web site it is, the ‘customer journey’ has always been important, otherwise you lose that engagement and the end result can hit the bottom line. That was always the case.

But in the modern era, customers expect an easy and simple experience, otherwise they move elsewhere. And with the enormous digitisation that’s taken place since COVID, the competition to be better is vast. Furthermore, the pandemic also created all sorts of new problems, because hybrid or remote teams now work at different times, leading to a spaghetti junction of data.

Startups like Milkymap (The Netherlands), Smaply (Austria) and Journeytrack (US) have appeared, attempting to address this burgeoning market, although it’s unclear if any of those have yet raised venture financing.

So in some sense the starting gun on this market has now been fired with the news that TheyDo has now raised a €12 million series A round is led by Blossom Capital, with participation from 20VC, also London-based. Participating in this round are were Angels such as Des Traynor (Intercom) Founder and Grisha Pavlotsky (Miro), as well as other senior angels from unicorns such as Figma, Snowflake, Calendly, Retool and Amplitude.

TheyDo’s ‘Journey Management’ platform tackles what’s known as ‘customer-centric alignment’, sector forecasted to be worth $48.5 billion by 2023 (according to Future Market Insights).

TheyDo purports to show businesses where all the incoming metrics and quantitative data is linked to, where the data is coming from, and how all that impacts the customer journey. This means that teams from CX to product, marketing, sales, and customer success have a more comprehensive ‘dashboard’ on the issues, in order to monitor and improve things.

Founded by Jochem van der Veer, Charles Beaumont and Martin Palamarz, TheyDo says it is now used by teams from Atlassian, Cisco, IBM, Johnson & Johnson, T-Mobile and Qualtrics.

Over a call, TheyDo cofounder Jochem van der Veer told me large businesses now put a premium on customer experience: “There is a mindset shift going on where these companies are now realising that organising around their customer journey is just the only way that they can stay relevant. And they are lacking tools to actually support that workflow.”

Warming to that theme, Harry Stebbings, founder of 20VC, added: “With the increasing consumer expectation of products there’s just a real fragmentation of roles within companies. When we look now at all the different roles within product from product management, product marketing, and everything in between, the fragmentation of all these different functions means that it’s harder and harder to create one great unified experience because you’re working in isolation. And so the specialisation of these roles means having one unified stream to create that great experience is more and more important.”

That’s where TheyDo comes in, says van der Veer: “It really is this whole way of organising large organisations internally, taking into account the customers perspective. And that is the massive trend going on that TheyDo is addressing.”

He says its customers now range from the Dutch Postal Service, to Johnson and Johnson, to Atlassian.

Interestingly, the journey towards the TheyDo product came out of van der Veer’s and his co-founders experience as an CX agency (or “SWAT team” as he calls it) where they would go into a company to figure out what was going wrong. Out of that experience they produced a product to ‘scratch their own itch’ and this eded up being TheyDo.

“We got hired as a consulting company into these large fortune 500 businesses to transform them from the inside. But we realised to scale our business we needed some software to to expand that we were doing. Eventually our customer would say ‘can we hire you to do this work and can we also get that tech solution you have?’ That’s when we realised there might be something bigger here,” he told me.

Stebbings points out that this is exactly a pattern he looks for in startups, as an investor: “I learned there’s a pattern when you have agencies that turn into products because they build them internally and their customers love them. From Intercom to MailChimp. They tend to work out very well.”

That remains to be seen in TheyDo’s case. However, it’s clear Blossom and 20VC are both counting on history repeating itself.

TheyDo fires the starting gun on the race to own the customer journey by Mike Butcher originally published on TechCrunch

For businesses, protecting trademarks is often a lengthy and expensive process, especially if they have multiple brands. Digip digitizes much of the process, helping its customers file trademarks by themselves instead of going to law firms. The Stockholm-based legaltech startup announced today it has added $1.3 million to its seed round, bringing the total to $3.4 million. The new funding was led by Industrifonden and Seed X, with participation from family offices and angel investors.

Founded in 2020, Digip now has 500 customers in 42 countries, ranging from startups to large enterprises that make hundreds of dollars in revenue. Digip currently makes about $500,000 in annual recurring revenue and that amount is forecasted to grow 3x during the 2022 fiscal year. Over the last year, Digip has also expanded its IP service into the United States and other international markets.

Co-founder and CEO Viktor Johansson told TechCrunch that Digip was founded after its team saw that entrepreneurs are reluctant to use traditional law firms that bill by the hour. To file trademarks, businesses usually ask a lawyer to conduct trademark searches. They are billed per search, which adds up quickly if a business has multiple brands they need to trademark. Then they have to pay for a lawyer to file trademark applications. But the process doesn’t end there. Businesses also have to monitor their trademarks in markets where they own it, and that is another charge.

Digip combines all these steps into one online workflow. Instead of charging for different parts of the process, its customers pay a flat monthly or yearly subscription fee, plus application fees charged by trademark offices.

Businesses can use Digip to research trademarks and get on-demand advice from its team for free. If they become subscribers, they can then use Digip’s platform to manage their trademark applications in 180 countries. The platform enables this with a trademark warehouse that has updated trademark data. Data collection and updates are automated as Digip enter new markets.

It also trains AI/ML algorithms for searches that cover 100 languages and manages customers’ trademarks by reading and interpreting trademark data. This enables onboarding to be automated and makes Digip’s process scalable.

Johansson said Digip initially considered offering its service to law firms, but decided not to since they are slow at adopting legal tech. But Digip does have a global network of lawyers that its customers can go to for support.

Johansson said that Digip’s largest markets are the United Kingdom, Nordic countries and the European Union, and it’s seeing more demand from the United States, Canada and Australia. Many of its customers are venture-backed businesses that run digital businesses in sectors in sectors including SaaS, deep tech, direct to consumer, life science, metaverse, blockchain and fintech.

“A cool thing with trademarking is that you pick up on early business trends,” Johansson said. “We have been involved in some interesting projects with emerging technologies that will hit the markets in coming years. This is a really fun and exciting part of our setup.”

Johansson said Digip’s closest competition are still law firms, but it also considers lawyers to be close collaborators. “Because digitization of legal has been slow many companies are stuck in legacy bills by the hour trademark solutions,” he said. “Some law firms that rely significantly on trademark filings are our competitors, whilst other law firms that do not do significant business through trademark filings see us as a great potential partner for them.”

Over the next few months, Digip will launch several new features. These include an open API that will let partners integrate Digip’s technology into their workflows. Johansson said users will see a significantly improved trademark search on Digip.com. The company will also expand into new markets over the next 12 months.

Digip digitizes the process of applying for trademarks by Catherine Shu originally published on TechCrunch

HR organizations are faced with a widening skills gap, economic headwinds and changing expectations around work. It’s no surprise, then, that burnout and exhaustion are widespread in HR, with one survey finding that 42% of teams are struggling under the weight of too many projects and responsibilities.

Change starts with personnel and management, some might argue. Others leaning more technoutopianist might proffer HR tech as a solution. While there’s a fair amount of dissatisfaction with HR tech vendors (at least according to some data), to be fair to the tech-positive crowd, many companies see real value in HR tech. According to a recent Sapient report, over half of businesses with more than 500 employees plan to increase HR tech spending by an average of 21% into the coming year.

One beneficiary of that increased spending is London-based Beamery, a startup developing a talent lifecycle management platform. Beamery today announced that it raised $50 million in a Series D round that values the company at $1 billion, bringing the company’s total raised to date to $228 million.

Teachers’ Ventures Growth (TVG), a part of the Ontario Teachers’ Pension Plan, led the round. “I believe Beamery is well-placed to win because it provides a solution that you can rely on through different economic cycles,” TVG’s Avid Larizadeh Duggan said in an emailed statement. “Beamery is helping the world’s largest employers with this talent agility, and allowing them to unlock the potential of their workforce.”

Certainly, Beamery gained impressive traction this year, growing the size of its customer base to “hundreds” of enterprises and over 25,000 users. Revenue from Fortune 500 clients rose by over 250% compared to June 2021, when Beamery closed its Series C round, according to the company, while net retention grew to 135%.

“Beamery’s … talent lifecycle management platform gives organizations, such as General Motors, VMWare and Johnson & Johnson, the intelligence they need to make the right decisions about their workforce and supports them through each stage of the talent lifecycle – from recruiting to talent mobility and development to upskilling,” Beamery CEO Abakar Saidov told TechCrunch in an email interview. “The new funding will support continued investment in our platform and tech capabilities and help to build out global sales footprint.”

Beamery was founded in 2013 by Saidov and his brother, Sultan Saidov, along with Mike Paterson. The Saidov brothers say their vision for Beamery had it origins in their experiences as children of immigrants, when they became aware of the structural challenges associated with work. Paterson was previously an analyst at Morgan Stanley, while the Saidov brothers worked at Goldman Sachs — Abakar as a commodities trader and Sultan as a mergers and acquisitions analyst.

Founded as Seed Jobs, Beamery uses AI to identify potential job candidate matches for open roles. Like many candidate-vacancy matching platforms, Abakar Saidov says that Beamery ranks skills based on the industry a company’s hiring for and a candidate’s relevant work experiences.

“Beamery uses [AI] in our talent lifecycle management platform to give companies the intelligence they need to plan for business needs and gaps, understand the skills and capabilities they have and attract, retain, upskill and redeploy their workforce successfully,” Abakar Saidov said. “[O]ur models are not meant to replace humans; instead, they give relevant information to human decision makers to make better decisions.”

Given the increased scrutiny over candidate-recommending AI systems, Abakar Saidov was quick to note that Beamery shows how various factors, including skills, seniority, proficiency and industry, influence its recommendations and to what degree. Beamery is among the vendors that could be subjected to a New York City regulation — the Automated Employment Decision Tools law, set to go into effect in January — that would ban employers from using AI hiring tools unless a bias audit can show they won’t discriminate.

Abakar Saidov says that Beamery recently completed a third-party audit for bias in its AI capabilities, which involved “rigorous testing” of the platform’s machine learning models. (Abakar Saidov didn’t proactively share a copy of the report with TechCrunch; we’ve requested one.) The company also partnered with Parity AI, a startup led by AI ethicist and activist Liz O’Sullivan, to audit the platform on an ongoing basis.

“Within the Beamery … platform itself (i.e. in the application layer), a key differentiator for us is helping customers ensure their own compliance with the myriad of global personal data and privacy standards,” Abakar Saidov said. “We achieve this primarily through the preference center, which lets candidates control their consent, whether and how companies can contact them and control how AI is used against their profile.”

Beamery doesn’t exist in a vacuum, of course. Competitors in the HR tech software space include 15Five, which raised $52 million in July for its talent management solution. There’s also Gloat, a well-capitalized startup building AI-powered internal jobs marketplaces. Eightfold is among the most formidable, with an over-$2-billion valuation and backing from SoftBank’s Vision Fund 2, General Catalyst and Lightspeed.

Broadly speaking, VCs have shown a willingness to put money behind HR tech startups even as other segments underperform. According to an analysis from WorkTech, the first half of this year saw the second-largest global work tech investment, surging to $9.4 billion, with $4.6 billion invested in Q2 alone.

Despite layoffs in the tech industry, job growth has remained resilient despite the economic headwinds, driving demand for HR tech — and spawning new vendors as a result.

To stay ahead, since its Series C, Beamery has doubled down on analytics capabilities, Abakar Saidov says — introducing a dashboard designed to enable companies to better understand their workforce by aggregating skills data across disparate HR systems and tools. The platform also recently rolled out a portal for candidates that provides recommendations for jobs as well as skills they might need to develop to further their careers in their chosen industry. And, as an outgrowth of its acquisition of internal HR sourcing platform Flux, Beamery launched Beamery Grow, which Abakar Saidov describes as a “talent marketplace solution” to help employees gain new skills and connections from within their organizations.

“We are prioritizing enhancements that will let customers quickly and easily leverage their talent data for things like agile workforce planning, as well as ensuring they have real-time intelligence and insights around their current and future workforce allocation, the skills that exist in their organization relative to their business outcomes and their achievement of diversity, equity and inclusion targets,” Abakar Saidov said. “The capabilities that a company will need over the next ten years are in many cases very different from today, and therefore HR tech solutions need to be able to help businesses build, buy or borrow the skills they need to build a future fit workforce.”

Beamery currently has 417 employees. When asked about hiring plans, Abakar Saidov said they’re “in development.”

Beamery, the all-in-one talent management platform, becomes a unicorn by Kyle Wiggers originally published on TechCrunch

Twitter today dispersed the Trust & Safety Council, which was an advisory group consisting of roughly 100 independent researchers and human rights activists. The group, formed in 2016, gave the social network input on different content and human rights-related issues such as the removal of Child Sexual Abuse Material (CSAM), suicide prevention, and online safety. This could have implications for Twitter’s global content moderation as the group consisted of experts around the world.

According to multiple reports, the council members received an email from Twitter on Monday saying that the council is “not the best structure” to get external insights into the company product and policy strategy. While the company said it will “continue to welcome” ideas from council members, there were no assurances about if they will be taken into consideration. Given that the advisory group designed to provide ideas was disbanded, it just feels like saying “thanks, but no thanks.”

A report from the Wall Street Journal notes that the email was sent an hour before the council had a scheduled meeting with Twitter staff, including the new head of trust and safety Ella Irwin, and senior public policy director Nick Pickles.

This development comes after three key members of the Trust & Safety council resigned last week. The members said in a letter that Elon Musk ignored the group despite claiming to focus on user safety on the platform.

“The establishment of the Council represented Twitter’s commitment to move away from a US-centric approach to user safety, stronger collaboration across regions, and the importance of having deeply experienced people on the safety team. That last commitment is no longer evident, givenTwitter’s recent statement that it will rely more heavily on automated content moderation. Algorithmic systems can only go so far in protecting users from ever-evolving abuse and hate speech before detectable patterns have developed,” it said.

After taking over Twitter, Musk said that he was going to form a new content moderation council with a “diverse set of views,” but there has been no development on that front. As my colleague, Taylor Hatmaker noted in her story in August, not having a robust set of content filtering systems can lead to harm to underrepresented groups like the LGBTQ community.

Twitter disperses the Trust & Safety Council after key members resigned by Ivan Mehta originally published on TechCrunch

China’s abrupt easing of its zero-COVID policy last week has led to a spike in cases and rising fears of the virus. Over the last three years, the authority has used big data to monitor the movement of people and thus control the spread of the virus. Now that the government is gradually undoing some of these tech-enabled restrictions, individuals are turning to private tech firms to manage the pandemic.

At the dawn of the pandemic in 2020, DXY, an online community for health professionals, swiftly introduced a fact-checking feature to fight COVID disinformation. But such grassroots efforts soon faded into the background as COVID cases remained rare in China and treatments happened at centralized, government-directed facilities.

The Chinese authority also rolled out a suite of COVID-prevention apps that became a digital pass for people to move around on a daily basis. These apps track individuals’ health status with diligent, compulsory COVID tests and monitor their potential exposure to the virus through travel history.

China is now doing away with some of these measures. As of Monday, the much-loathed travel tracing app was dropped, providing some relief to those who are wary of the tool being abused to control one’s life. Major cities like Guangzhou and Beijing have advised patients with mild or no COVID symptoms to isolate at home, ending a nearly three-year-long practice wherein those infected with COVID were sent to makeshift quarantine hospitals regardless of their symptoms.

As the virus is expected to spread in the upcoming weeks with people left to their own devices, Chinese tech firms are coming forth with initiatives to help navigate a new wave of infections.

Over the past week, more than a million users have sought medical advice from doctors remotely through JD Health, the healthcare arm of Chinese ecommerce giant JD.com, a company spokesperson told TechCrunch. The types of remote healthcare the platform provides include advice on COVID prevention, chronic disease management, recovery plans, and psychological counseling. Asymptomatic and those with mild symptoms are also able to get prescriptions via JD Health.

JD Health introduced the online COVID clinic shortly after China’s snap announcement to phase out some of its most draconian COVID policies. Baidu’s map began showing in real-time the stock status of antigen test kits at local pharmacies, of which demand has surged after China removed compulsory, heavily-subsidized nucleic acid tests. According to JD’s online marketplace, the transaction volume of rapid test kits rose 344% week over week on December 10.

Partly thanks to China’s persistence on zero-COVID, some of the infrastructure for infection prevention is already in place. Smartphone-enabled contactless ordering at restaurants is already a norm across the country. Pickup kiosks that temporarily store food deliveries are also a common sight, doing away with the need for customers and couriers to meet in person.

Some sectors are less resilient to COVID impact. The logistics industry is particularly under pressure as couriers are expected to get hit by the virus like the rest of society, while the demand for express delivery climbs as people rush to hoard medications and isolated individuals rely on grocery delivery. Factory bosses who worry that outbreaks would shut down manufacturing have invested in robots, but many of them are finding the costs of upgrading production lines too high in the short run, some of them told TechCrunch.

As China relaxes zero-COVID, tech firms assume a larger role in fighting the virus by Rita Liao originally published on TechCrunch

Car parts buyers require specific parts to fit specific vehicles, making for a supply-constrained environment. New Zealand-based Partly wants to ease those constraints by connecting parts buyers around the world with the correct parts.

The two-year-old startup is not a car parts marketplace. Rather, Partly powers marketplaces like eBay and Shopify with its database of over 50 million parts from over 20,000 suppliers and OEMs.

“The way the tech works in principle is we work with the suppliers to ingest, structure and standardize all the data,” co-founder and CEO Levi Fawcett told TechCrunch.

Then the company manages that data and pushes it back onto big platforms that buyers are already using to find car parts.

The startup on Monday closed a $21 million Series A to continue growing in Europe, where the majority of its customer base is — aside from marketplaces like eBay, Partly also works with the United Nations and a couple of unnamed Fortune 500 companies. The startup also aims to use the funds to scale more aggressively in the U.S., where it’s actively hiring and building an office. Most importantly, the funds will help Partly double its engineering team to work on the core problem of aggregating all the correct parts of a vehicle just based on a license plate.

“Sounds simple, but it’s a ridiculously hard problem,” said Fawcett, who noted Partly’s team of 50 should cap out over 100 staffers by the end of next year.

A secondary goal for Partly, besides scaling its business, is to represent New Zealand on the world stage. With top tier clientele and no direct competitors, the startup aims to be the largest NZ-based tech firm within five years. To do that, it’ll have to contend with Xero, which is publicly traded on the Australian Stock Exchange and has a market cap of around $7.4 billion, per Google Finance data.

Fawcett, who previously managed and developed hardware simulations at Rocket Lab, said the opportunity to connect part buyers with the correct parts is “monstrous.” In the U.S. alone, consumers spent close to $95.4 billion on motor vehicle parts and accessories in 2021. The auto parts and accessories market is expected to reach a global market size of $2.5 trillion by 2024.

“About 98% of parts ordered today is done on the phone by a parts interpreter, and it’s their job to take the phone call, understand what they’re looking for, find it in the system, figure out what vehicle it’s come from, decide if there are any differences or if it was modified when it came from another country, and then provide the buyer with the right part,” said Fawcett. “It’s that whole process we’re flipping. Instead, you put in your license plate and then pick the part you want. It’s basically taking a super archaic process and radically changing it by removing the human.”

The problem hasn’t been solved at scale before because it requires working across vehicle manufacturers, aftermarket part manufacturers and retailers, and building a common language so all the information across manufacturers is consistent. This not only makes it easier for buyers, but also for sellers that want to better understand their customers.

Ex-Rocket Lab engineer raises $21M for Partly to make buying car parts easier by Rebecca Bellan originally published on TechCrunch