Privacy watchdogs in the U.K. and Canada have launched a joint investigation into the data breach at 23andMe last year.
In its data breach notices, the company said it didn’t detect the hackers’ activities for around five months, from April until September 2023.
23andMe said it only became aware of the account breaches in October 2023, when hackers advertised the stolen data on the unofficial 23andMe subreddit and a well-known hacking forum.
Hackers broke into around 14,000 accounts of 23andMe customers by reusing their passwords from previous breaches, a technique known as password spraying.
That’s how the hackers were able to scrape information on 6.9 million users by only hacking 14,000 accounts.
When you’re looking for a startup idea that could slow climate change, you might become an expert at home energy assessments.
Instead, the startup has put together a small team of engineers to create its own AI model specialized in home energy assessments using machine learning.
The company uses open data, such as satellite images, as well as its own training data set with millions of photos and energy assessments.
In the company’s first tests, its home energy assessments have been accurate within 5% of old-fashioned assessments.
The startup has now raised €4.7 million ($5.1 million at today’s exchange rate) with Racine² leading the round and a non-dilutive investment from Bpifrance.
It’s the first time that the number of affected Snowflake customers has been disclosed since the account hacks began in April.
So far, only Ticketmaster and LendingTree have confirmed data thefts where their stolen data was hosted on Snowflake.
Several other Snowflake customers say they are currently investigating possible data thefts from their Snowflake environments.
Mandiant said the threat campaign is “ongoing,” suggesting the number of Snowflake corporate customers reporting data thefts may rise.
Last week, TechCrunch found circulating online hundreds of Snowflake customer credentials stolen by malware that infected the computers of staffers who have access to their employer’s Snowflake environment.
Human Native AI is a London-based startup building a marketplace to broker such deals between the many companies building LLM projects and those willing to license data to them.
Human Native AI also helps rights holders prepare and price their content and monitors for any copyright infringements.
Human Native AI takes a cut of each deal and charges AI companies for its transaction and monitoring services.
Human Native AI announced a £2.8 million seed round led by LocalGlobe and Mercuri, two British micro VCs, this week.
It is also a smart time for Human Native AI to launch.
Snowflake’s security problems following a recent spate of customer data thefts are, for want of a better word, snowballing.
TechCrunch earlier this week found online hundreds of Snowflake customer credentials stolen by password-stealing malware that infected the computers of employees who have access to their employer’s Snowflake environment.
It’s not yet known how many Snowflake customers are affected, or if Snowflake knows yet.
Snowflake said it has to date notified a “limited number of Snowflake customers” who the company believes may have been affected.
Snowflake declined to say what role, if any, the then-Snowflake employee’s demo account has on the recent customer breaches.
LinkedIn has confirmed it will no longer allow advertisers to target users based on data gleaned from their participation in LinkedIn Groups.
In response to the complaint it received in February, the EC wrote to LinkedIn to request further information on how it might be enabling targeted ads based on sensitive personal data such as race, political allegiances, or sexual orientation.
While LinkedIn maintained that it complied with the DSA, the company has now removed the ability for advertisers to “create an advertising audience” in Europe using LinkedIn Group membership data.
“We made this change to prevent any misconception that ads to European members could be indirectly targeted based on special categories of data or related profiling categories,” Corrigan wrote on LinkedIn today.
LinkedIn will still allow targeted advertising, just not using data garnered from LinkedIn groups.
Two senior police officials are accused of sharing citizens’ personal information from a classified government database with criminalsTwo senior officials working for anti-terror police in Bangladesh allegedly collected and sold classified and personal information of citizens to criminals on Telegram, TechCrunch has learned.
According to the letter, the police agents were caught after investigators analyzed logs of the NTMC’s systems and how often the two accessed it.
Last year, a security researcher found that the NTMC was leaking people’s personal information on an unsecured server.
Another Bangladeshi government agency, the Office of the Registrar General, Birth & Death Registration, also leaked citizens’ sensitive data last year, as TechCrunch reported at the time.
Although the incident is under investigation, a well-placed source within the government told TechCrunch that there are still officials who are offering to sell citizens’ data.
BIRA also takes aim at Amazon’s “Buy Box,” claiming that Amazon manipulated which retailers were selected for the coveted placement.
BIRA and its legal team say that the claim is the biggest ever collective action to be launched by retailers in the country.
Making itself a must use for retailers, Amazon has then proceeded to cause damage and financial loss to retailers by misusing their confidential data that Amazon was entrusted to keep safe and by preferencing its own retail operations.
Retailers in the U.K. were entitled to be treated better and fairly by Amazon.
The U.K. is its largest international market, where it made $33.6 billion in revenues in 2023 (out of $575 billion in global revenues).
But as the data analytics and AI boom drives organizations to expect more of data models, many of the old paradigms are proving difficult to manage — and exceptionally brittle.
Now, five years later, Keydunov and Tiunov have a veritable business on their hands, having launched a subscription-based service built on Cube — Cube Cloud — that adds automated workflows and enterprise-focused governance and deployment tooling.
An illustration of Cube’s semantic data layer.
Image Credits: Cube“Cube Cloud is a universal semantic layer that is an independent, yet interoperable, part of the modern data stack that sits between your data sources and data consumers,” Keydunov said.
Keydunov says that the open source Cube project has surpassed 10 million downloads, while Cube Cloud is now installed on around 90,000 servers.
Not all generative AI models are created equal, particularly when it comes to how they treat polarizing subject matter.
They found that the models tended to answer questions inconsistently, which reflects biases embedded in the data used to train the models, they say.
“Our research shows significant variation in the values conveyed by model responses, depending on culture and language.”Text-analyzing models, like all generative AI models, are statistical probability machines.
Instrumental to an AI model’s training data are annotations, or labels that enable the model to associate specific concepts with specific data (e.g.
Other studies have examined the deeply ingrained political, racial, ethnic, gender and ableist biases in generative AI models — many of which cut across languages, countries and dialects.
