On Monday evening, Greek Prime Minister Kyriakos Mitsotakis arrived in San Francisco to advocate for Greece, engaging in an activity not commonly practiced by state leaders on trade trips. Speaking to an audience of around 250 founders, investors, and operators, he candidly admitted that he lacks answers to numerous AI-related queries that leaders globally are discussing more discreetly.

During our conversation at an event organized by Endeavor Greece, the local branch of a global nonprofit supporting entrepreneurs in developing and growth economies, the prime minister characterized his visit to the Bay Area as, in part, a mission to gather information. He spent the morning visiting Tesla and Sequoia Capital, among other locations, while planning to proceed to this week’s U.N. General Assembly in New York. However, the journey also aimed to foster connections; indeed, he was appealing to anyone in the tech sector who might consider establishing a presence in Greece, which is expected to regain its developed market status from MSCI next year, an influential index provider.

“This reflects another sign that the economy is performing well and that Greece is no longer seen as an exceptional case,” Mitsotakis remarked.

Calling the trip “a homecoming,” Mitsotakis, who holds a master’s degree from Stanford, emphasized the economic narrative during our discussion. Greece is reducing its debt at an unprecedented rate and is also borrowing at lower rates than the United States. “I’m not sure if I should mention this,” he said with a smile to the audience.

This isn’t a like-for-like comparison. The eurozone has lower interest rates, which partially explains the discrepancy. However, it serves as a compelling point for discussion. Currently, Greece’s 10-year bond yield hovers around 4.3%, compared to roughly 5% for U.S. Treasuries. This was entirely unthinkable during the debt crisis, when Greek yields soared past 40% in 2012.

Mitsotakis was keen to discuss Greece’s financial expenditures as well. He mentioned that a significant portion of the approximately €36 billion Greece acquired from the EU’s post-COVID recovery fund was invested in digital infrastructure, which includes an online platform allowing Greeks to manage government paperwork without queuing, and a new supercomputer developed in collaboration with Hewlett Packard Enterprise in Lavrio, which he indicated would be operational within months to enhance AI and scientific research in Greece.

As expected, he also highlighted policy changes that could be even more beneficial to tech companies. Greece has restructured the taxation of stock options, relaxed labor regulations, and provides returning Greeks much lower tax rates for up to seven years.

Additionally, we discussed the nation’s expanding range of visa options, although when queried about how he gauges the success of these, Mitsotakis indicated that the government “still [has] work to do” regarding processing speed.

He also noted a remarkable transformation in Greece’s public universities, which he claimed previously adopted a “radical leftist approach,” disparaging corporations, but are now fostering startup initiatives.

Mitsotakis aims to reclaim the talent Greece lost during its debt crisis when many young citizens departed due to a lack of options. He also recognizes an opportunity: as obtaining American work visas becomes more challenging, he encouraged more founders to contemplate hiring talent in Greece.

I noticed a shift in tone when the discussion turned to the societal impacts of AI. Mitsotakis was refreshingly open about his lack of a comprehensive strategy.

For instance, Greece is planning a ban on social media for children under 15 starting this January, a measure that many nations are also considering or implementing. However, he remarked that this ban might already be outdated, given the seemingly addictive qualities of AI chatbots.

“At times, I feel as though we’re fighting a battle from the past,” he noted. “What implications does it have for our children to grow up with digital companions or romantic partners?”

He expressed similar reservations regarding AI in educational settings. Greece has been piloting an education initiative with OpenAI aimed at alleviating teachers’ administrative burdens, and he sees potential in personalized AI tutors. Nonetheless, he emphasized that these advantages are contingent upon AI complementing rather than substituting “the hard work of mastering fundamental skills.” He cautioned that students are already resorting to chatbots for homework assistance, stating: “complacency is a human characteristic.”

When I inquired about AI infrastructure — Greece is actively seeking data center investments — he characterized the nation as an exception. In contrast to growing global opposition to data centers due to their significant demand for electricity and water, Mitsotakis commented that “we have not encountered any notable backlash” within Greece. He highlighted that Microsoft is constructing a cluster of data centers near Athens, and that AWS recently publicized an agreement with Greece’s largest utility to develop the country’s most extensive data center in a former coal area. According to him, these projects are received positively.

Differing from many politicians, he didn’t feign knowledge regarding future developments. Job displacement “is an occurrence that is going to take place,” he insisted, “and no government or society is equipped for the rapid pace at which it will unfold.”

Regarding the discussion on whether AI development should be slowed, he aligned himself with the frontier lab leaders advocating for that stance. “If those who create the models are expressing uncertainty about their functionality,” particularly regarding their self-enhancement, he commented, “we must heed their warnings.”

Some form of “intelligent regulation” is unavoidable, he noted, and the U.S. will primarily influence its design. Nevertheless, he added, Greece would be eager to facilitate that dialogue.

Humanity has generated “a kind of intelligence that is rapidly exceeding the capabilities of the most sophisticated organ created through evolution,” Mitsotakis stated, raising profound questions about human existence. He proposed uniting technologists with “social scientists, philosophers, and historians,” asserting, “I can’t imagine a better location to engage in these discussions.”

He has a valid point. Athens is where Socrates engaged his fellow citizens in the agora, the historic public square, discussing the essence of a good life and a just society. Perhaps the debates still occurring in Silicon Valley will also find resolution there?

Beginning Wednesday, Discord, the messaging platform for communities, will implement age verification protocols for all users.

This initiative follows the company’s prior postponement of age verification plans in February due to backlash from users worried that biometric scanning and ID verification could threaten their privacy. Those concerns were valid; last year, a third-party vendor collaborating with Discord experienced a breach, potentially compromising the data of 70,000 users, including government ID images and selfies used for identity verification.

“Age assurance invites strong opinions and skepticism, particularly when it involves disclosing biometric data or a type of ID,” stated Discord CTO Stanislav Vishnevskiy in a blog post published on Tuesday. “Simultaneously, age assurance regulations are evolving, and since my previous post, we’ve had to initiate age assurance in additional regions, including Brazil and Texas.”

Nonetheless, Vishnevskiy mentions that over 90% of Discord users will not need to undergo age verification. He clarifies that Discord will utilize existing account information to identify “signals,” such as account longevity (the duration an account has been active), device metadata, activity data, and other usage indicators, to categorize users into adult, teen, or unverified age brackets.

Discord asserts that it does not deduce users’ ages based on message content, profile information, or demographic characteristics.

“The manner in which a typical 15-year-old interacts on Discord (including the number of communities they join and the way they engage with others) differs substantially from that of a typical 30-year-old,” the company outlined in a separate technical blog post published on Tuesday. “Our [machine learning] model has recognized these patterns to assess if a particular account likely pertains to an adult or a teen.”

This does not indicate that Discord is acquiring any new user data. It is commonplace for social media entities to retain this kind of information about their users, although it may feel discomforting to see it expressed so explicitly.

Due to local laws necessitating Discord to employ age assurance technology in specific areas, the company could educate its model on accurately verified data linking user actions to age.

If Discord lacks confidence in a user’s adult status, that user can still verify their age without undergoing biometric or ID scanning. They can appeal by providing credit card details or age-related information from their Apple App Store or Google Play Store accounts.

For confirmed adults, the platform will remain unaffected. However, teen accounts will automatically be restricted from entering age-limited servers or channels. Message requests from individuals who aren’t their friends will redirect to a distinct message requests inbox, and teens will receive notifications upon accepting friend requests from someone they don’t mutually share friends with.

As governments globally persist in enacting age verification mandates to safeguard children online, experts caution that these age checks could introduce various new privacy and security challenges. This predicament presents platforms like Discord with a tough decision: comply or face penalties.

On November 4, the Founder Summit by TechCrunch will host an essential one-day intensive on building startups at Boston’s SoWa Power Station. Founders shouldn’t have to endure the toughest lessons in the toughest ways, and this event aims to simplify the challenges of launching a business and amplify the rewards.

Rather than spending months on trial and error, you can gain direct insights from the investors and founders who have faced the decisions you’re now navigating. We’re covering everything from securing funding and recruiting to AI strategy and dominating your sector.

Don’t just take our word for it. Check out the complete lineup of speakers and sessions we have arranged for the Founder Summit below, and secure your spot for the event. All tickets are offered at the most affordable rates available, making this the ideal moment to take action.

The TechCrunch Founder Summit Official Agenda Unveiled

Without any more delay, here’s the official agenda. Familiarize yourself with every session and speaker listed on the event agenda page.

The New Guidelines for Fundraising 

Securing funds has never been simple, yet the strategies continue to evolve. Brian Devaney, partner at Underscore, will clarify what investors currently desire, how founders can distinguish themselves in a saturated market, and the areas where founders frequently lose influence without being aware. Expect a straightforward examination of today’s fundraising landscape, covering everything from initial investments to term agreements, and the blunders that can determine the fate of a funding round.

The CEO Role Is Ever-Challenging 

While the job title remains constant, the responsibilities shift with every phase of a company’s expansion. HubSpot co-founder and Sequoia partner Brian Halligan will discuss the insights he has gained from mentoring and supporting startups through periods of swift growth, difficult choices, and continuous adaptation.  

This session will delve into how founders can transition into effective CEOs, sidestep typical leadership mistakes, and create businesses that will succeed long after their initial launch.

Designed for AI from the Start 

Integrating AI features is one thing; establishing an AI-centric business is an entirely different endeavor. Lior Div, co-founder and CEO of 7AI, will assess what adjustments must be made when AI is integrated as the core of a startup rather than just an addition. This session will explore everything from team organization and product innovation to operations and market entry strategies, examining what founders need to reconsider while operating in the AI age.

Not All Dollars Are Created Equal 

Raising funds is frequently viewed as a key achievement, but selecting the appropriate investors can influence a company’s trajectory long after the funds are received. Following an $11 million funding round for Cogent Security, co-founder and CEO Vineet Edupuganti gained firsthand knowledge of what differentiates valuable partners from costly distractions. This session will discuss how founders should assess investors, manage compromises, and think past valuation when constructing their investment table.

Identifying the ‘King of the Hill’ Company — Hidden Indicators That Define Market Leaders 

Innovation is essential for addressing the world’s greatest challenges, yet only select companies are likely to evolve into market-defining leaders on a large scale. In this session, TDK Ventures investment director Tina Tosukhowong unpacks the firm’s “King of the Hill” model: the rigorous framework the top investors use to assess companies based on their economics, scalability, and market timing.  

Using real-world examples, including Tina’s insights into fission and fusion, attendees will gain actionable frameworks for assessing startup readiness, charting competitive landscapes, and recognizing when timing, talent, and technology come together.

Your Company Reflects Your Hiring Choices 

Initial hiring decisions can shape a startup’s culture, pace, and execution capabilities. Melissa Taunton, partner at NEA, will offer insights gained from working with founders as they form teams amid rapid growth and ambiguity.  

From pinpointing the right initial hires to dodging common recruitment pitfalls, this session examines how founders can develop organizations that are robust, flexible, and ready for whatever lies ahead.

The Boston Founder’s Guide 

Silicon Valley often attracts the spotlight, but Chase Garbarino has been steadily creating category-defining companies out of Boston for over ten years. As the co-founder and CEO of HqO, he has secured $200 million from investors and expanded to over 30 countries without ever needing to shift locations.  

He will share insights into the genuine advantages and challenges of operating outside the traditional startup landscape, as well as what every founder outside a major entrepreneurial hub needs to know.

Discovering Product-Market Fit Before Scaling 

Every founder aspires to achieve product-market fit, but recognizing when you’ve truly found it is considerably more complex. Kent Bennett, partner at Bessemer Venture Partners, will clarify how founders can verify demand, create a minimum viable product that addresses key queries, and avoid premature scaling before essential elements are established.  

This session will delve into the vital signals, misleading metrics, and choices that differentiate sustainable companies from costly trials.

What Awaits Beyond the Sessions of the Founder Summit 

The programming constitutes a significant aspect of TechCrunch Founder Summit’s value, but it’s not the entire picture. You’ll also have the opportunity to connect with fellow founders tackling similar fundraising and recruitment challenges, and network with individuals who have already navigated those paths. Or even better…those who have financed those who have endured the startup journey.  

A day filled with interactive discussions transforms into months of actionable insights at TechCrunch Founder Summit, and we look forward to having you join us and the broader Boston startup ecosystem on November 4. Register now to seize your ticket savings and participate in the ultimate founder’s bootcamp.

Snorkel AI, a company assisting AI laboratories and businesses in creating training datasets and simulated environments, has secured $350 million in a Series E funding round, achieving a valuation of $3.5 billion.

This latest funding round, spearheaded by Insight Partners and S32, values the seven-year-old company at almost three times the $1.3 billion valuation it received after raising $100 million in a Series D round 17 months prior. Current investors, including Addition, Lightspeed, Greylock, GV, and Wells Fargo, also took part in this round.

Initially, Snorkel offered software for automating data-labeling processes, but last year it transitioned to delivering finalized datasets to its customers, branding this service as data-as-a-service. Instead of functioning solely as a marketplace for human expertise, Snorkel employs a combined strategy, employing its software and models to create data synthetically along with input from subject matter experts.

According to Snorkel, its present annualized revenue run rate has reached $375 million, representing an 18-fold increase over the past year. This surge is driven by the unquenchable need for high-quality training data from AI labs.

Other data firms positioning themselves as AI data laboratories are experiencing a similar surge in growth. Mercor has seen its gross annualized revenue grow to $2 billion, Handshake surpassed the $1 billion mark earlier this year, and TechCrunch has reported that Micro1 has expanded to $500 million. Since these companies allocate approximately 60% to 70% of their gross revenue directly to the domain experts performing the tasks, it is crucial to understand that their actual net annual revenue is significantly lower than those reported gross figures.

Given that Snorkel offers reinforcement learning (RL) environments and full datasets instead of human labor, payments to its human experts are incorporated into its cost of goods sold rather than reported annualized revenue figures, as stated by the company.

Snorkel began commercial operations in 2019 after four years of research by co-founder and CEO Alex Ratner and his team at a Stanford AI lab.

Today, Qualcomm unveiled two new flagship mobile processors named Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6, emphasizing enhancements for AI-centric functionalities.

During its yearly Snapdragon Summit, Qualcomm highlighted that the latest chips could support enhanced personalization for AI assistants. These processors feature innovative sensing hubs capable of handling compact models with up to 200 million parameters. Thanks to the new sensing hub, smartphones can locally operate a personal assistant and recognize different speakers. Additionally, it can generate user-based memory to improve suggestions for task automation. The company mentioned it could support a complete voice-in and voice-out assistant via the new chip.

The Snapdragon 8 Elite Gen 6 incorporates a new accelerator designed for more effective model execution. The high-end Extreme variant can execute a mixture-of-experts (MoE) model with 30 billion parameters on-site. This implies that although the total model comprises 30B, it activates only a select number of parameters for specific tasks.

In contrast, at its Worldwide Developer Conference (WWDC) in June, Apple unveiled a mixture-of-experts model with 20 billion parameters, representing the pinnacle of its third generation of foundation models.

The new Qualcomm CPU also offers pixel-level control for camera systems, enhancing professional-grade experiences, alongside refined stabilization and motion interpretation. The Extreme variant is capable of 8K video recording at 60fps and 4K240 for ultra HD slow-motion. It also introduces the Advanced Professional Video (APV) codec for high-quality recordings.

Furthermore, both chips leverage AI technology to enhance vocal clarity and minimize background noise. It also features Qualcomm’s innovative voice bubble technology, which isolates the user’s voice during conversations.

At the event, Motorola revealed its Motorola Signature 27 smartphone, powered by the Snapdragon 8 Elite Extreme Gen 6, with general release expected later this year.

Qualcomm has been developing over 40 AI devices, yet there remains a consensus that many individuals will prefer using their phones for AI applications instead of separate devices. Nothing co-founder Carl Pei has previously indicated this trend, and more recently, new Apple CEO John Ternus echoed the same sentiment during the iPhone Duo introduction earlier this month.

Initial users of Meta’s Muse have been pondering that the reason for its impressive performance is its underlying foundation in OpenClaw, presented in a more user-friendly format. Meta has acknowledged that these observations are not entirely unfounded. As per an X post by Nat Friedman, head of product at Meta’s Superintelligence Labs (MSL), Muse was “certainly significantly inspired as a product by OpenClaw.”

Nevertheless, he clarified that Muse itself was “developed from the ground up.”

Friedman, the previous GitHub CEO who joined Meta last year together with Meta’s Chief AI Officer Alexandr Wang, mentioned that the Meta team had become enamored with OpenClaw and aimed to create Muse as “something akin to OpenClaw” capable of scaling to billions of users.

Given that OpenClaw is an open-source initiative, it isn’t surprising that Meta sought inspiration from it, particularly because of its rapid rise, which led OpenAI to acquire its creator earlier this year. However, this aligns with Meta’s established strategy of taking promising innovations and replicating their most effective features — a practice it notably executed with Snapchat’s creation of the stories format.

Friedman’s comment on X came in reaction to a viral post among the AI community, where Ansh Nanda, a co-founder of an AI app, claimed that “Muse is LITERALLY OpenClaw for the average person.”

Nanda’s X post featured a dialogue with Muse, in which the AI agent indicated that the resemblance between its own system files and those from OpenClaw was not merely “a coincidence” but actually “a match.”

Shortly thereafter, this discussion (and X in general) began to gain traction as others shared their own observations and analyses. For example, one user highlighted that Muse also contained a SOUL.md file, the straightforward text configuration file (written in a basic formatting language known as Markdown) that outlines an AI agent’s personality, interaction style and tone, values, limitations, and areas of expertise.

Another user observed that the files were not only identically named between the two assistants; their content was nearly identical as well.

In the wake of his X post, Friedman also addressed a query regarding why Muse had replicated the exact file names characterizing the AI agent’s workspace and had “virtually indistinguishable content” for the SOUL.md file. Rather than contesting these accusations, Friedman simply stated that “we believed Peter [Steinberger, creator of OpenClaw] had those aspects spot on.”

“We constructed Muse from the ground up, but it is unquestionably heavily influenced as a product by OpenClaw. After I experienced OpenClaw in January, I procured hundreds of Mac minis for the MSL team, and many of us became captivated by utilizing OpenClaw (and other personal agents),” Friedman stated on X. (TechCrunch has amended Friedman’s punctuation for clarity.)

“@steipete [Peter Steinberger] is a visionary, and his framework was groundbreaking from the very beginning. I believe many individuals drew inspiration from it. Our aim with Muse was to create something like OpenClaw that we could ensure is safe, secure, user-friendly, and scalable to billions of users,” he added.

The Muse app has proven to be quite successful thus far, having recently achieved the top position on the U.S. App Store. Data suggest that it is now surpassing ChatGPT’s launch when comparing platform and market availability directly.

When contacted for a statement, a Meta spokesperson referred to Friedman’s remarks, stating the company had no further comments beyond that.

ShinyHunters, a notorious hacking group recognized for extensive data breaches and blackmail, claims to have infiltrated the FBI and exfiltrated information on thousands of agents and applicants. The hackers announced this on their dark web leak platform, which TechCrunch examined, asserting they had acquired “confidential data about nearly all FBI agents and individuals who have submitted job applications to the FBI.”

404 Media was the first to report the breach after obtaining a sample of the compromised names, residential addresses, and phone numbers of FBI agents and their spouses, confirming part of the stolen information against publicly available records. The hackers declare their intrusion is “not financially driven,” and are insisting that the FBI retract a report they allege contains misleading accusations about them.

The independent outlet indicated that the hackers breached an Oracle PeopleSoft server, commonly utilized by HR departments and recruiters to maintain job applicants’ personal data, then transitioned to compromise an Amazon-hosted government cloud that holds the data of agents and applicants.

ShinyHunters informed the publication that they extracted terabytes of data but did not specify their intentions for the information if the FBI fails to remove its public report. The unauthorized data could pose a significant counterintelligence risk, where hackers and foreign operatives leverage the information to manipulate or blackmail FBI agents and their families into aiding a foreign nation.

The hackers allegedly vandalized the FBI’s jobs portal, which at the time of publication indicated that the site was “currently down for maintenance.” It also stated that the FBI’s special agent applicant site was likewise unavailable.

An FBI representative informed TechCrunch: “The FBI is aware of assertions regarding unauthorized actions affecting FBIjobs.gov and is presently conducting an investigation.” When contacted by TechCrunch, the ShinyHunters hackers did not disclose the number of individuals whose information was compromised but claimed they are “very confident we have data on mostly all of FBI,” and that “a significant amount of applicant data [is] involved as well.”

This marks the second confirmed breach of an FBI system this year after unidentified hackers accessed one of the agency’s systems for handling real-time wiretaps and foreign intelligence warrants, which could have pinpointed the agency’s surveillance targets. Additionally, FBI director Kash Patel’s personal email account was compromised and exposed by an Iran-aligned hacking group named Handala in reprisal for U.S.-led attacks against Iran.

Updated with remarks from FBI and ShinyHunters.