Techin' Geek

Over three years after the introduction of ChatGPT popularized generative AI, OpenAI is expanding its target audience from individual users to families.

OpenAI is looking for a specialized product manager in San Francisco tasked with creating experiences for families, caregivers, and older adults across its offerings. The position requires a background in developing products aimed at parents and families, as well as other consumer experiences sensitive to trust, based on the job listing.

This recruitment coincides with ChatGPT’s user base expanding beyond just younger individuals. Estimates from Sensor Tower, shared exclusively with TechCrunch, show that the proportion of global ChatGPT users aged 35 and older increased to 31% in Q2, up from 26% a year prior, while users aged 18 to 24 decreased from 34% to 29%. In the United States, nearly one in four smartphone parents utilized ChatGPT during the quarter, rising from 16% in the previous year, according to the firm’s estimates.

OpenAI did not reply to inquiries regarding the job listing.

The creation of a product role centered on families indicates that OpenAI is starting to view its offerings less as tools for solitary productivity and more as technology tailored for households, remarked Ben Bajarin, CEO of the technology consultancy Creative Strategies.

“This mirrors the trajectory that Google, Apple, and Meta eventually followed as their platforms became woven into daily life, but AI elevates the stakes since the assistant not only mediates content or devices,” he stated to TechCrunch.

This transition also presents new challenges related to trust and safety. Stephen Balkam, the chief executive of the Family Online Safety Institute, noted that the hiring reflects both the evolution of OpenAI and an increasing acknowledgment that AI products utilized by minors require distinct protections compared to those created for adults.

“I view this as safety through redesign,” Balkam shared with TechCrunch. “You take the initial product or service that was released… not really considering kids… so this is a much-needed reaction and response.”

These remarks come as new research released this week by the Family Online Safety Institute discovered that parents are underestimating how frequently their children use generative AI. While 27% of U.S. parents indicated their child had used generative AI in the past week, 38% of children reported using it themselves, according to a survey of over 4,000 families in the U.S. and Australia.

Balkam told TechCrunch that AI firms should develop products with younger users in mind, implementing stronger content controls, age-appropriate experiences, parental oversight, and reminders to make it clear users are interacting with an AI — not a human.

The recruitment also arises amidst increasing scrutiny regarding how AI companies safeguard younger users. OpenAI has faced several lawsuits from parents alleging that ChatGPT has contributed to harm experienced by their children, including incidents linked to suicide.

In addressing some of these worries, OpenAI has implemented a range of safety initiatives over the past year, including parental controls for teenage accounts, directing sensitive dialogues to reasoning models intended to better manage signs of distress, and, more recently, an optional “Trusted Contact” feature that can notify a family member or caregiver in potential self-harm situations.

AI companies, according to Balkam, have a chance to learn from the errors made by social media platforms, which for years treated children similarly to adults before enhancing their safeguards under growing public and regulatory pressure.

The hiring also supports OpenAI’s extensive initiatives surrounding families. During a recent workshop in collaboration with the San Antonio Spurs Community Impact organization and the Positive Coaching Alliance, the company expressed its goal to explore AI’s role in education, coaching, and youth engagement.

Nevertheless, the demographic transition is not exclusive to ChatGPT, even though OpenAI’s audience is evolving in some unique ways.

Sensor Tower estimates that individuals aged 25 to 34 make up 40% of the total app user base for Anthropic’s Claude and Google’s Gemini, similar to ChatGPT, contrasting with 33% for Microsoft’s Copilot. Copilot, however, skews older, with 20% of its users aged 45 and up, compared to 14% for Claude, 12% for Gemini, and 11% for ChatGPT.

While ChatGPT remains relatively less popular among older users, its growth rate is surpassing that of its competitors. The percentage of users aged 45 and over climbed by three percentage points year-over-year in the second quarter, contrasting with a two-point rise for Copilot and declines for Claude and Gemini, based on Sensor Tower’s analysis.

Among U.S. smartphone parents, Gemini demonstrated the broadest reach at 32% in Q2, followed by ChatGPT at 24%, Claude at 4%, and Copilot at 2%.

For Bajarin, the decision by OpenAI to recruit a product manager focused on families indicates the future direction of consumer AI. As AI evolves into technology utilized across different generations, he anticipates companies will introduce family plans, child and teen profiles, caregiver tools, shared household memories, AI tutoring, and enhanced safety measures.

The U.S. federal cybersecurity organization CISA stated that it lacked a pre-established response strategy for addressing a cybersecurity event in May, after an investigative journalist informed them that a contractor had inadvertently revealed sensitive keys and access credentials for U.S. government systems.

CISA, the Homeland Security division responsible for protecting federal networks and assisting in the defense of critical infrastructure, disclosed on Friday in a post-incident report that its personnel “had to allocate time to develop [a playbook] in the initial phases of the incident.” The agency emphasized the necessity of creating playbooks for “all expected requirements” to guarantee that organizations are prepared to react to a security breach instead of hurriedly crafting one in real-time.

The agency did not specify the extent of the delay in CISA’s response due to the absent playbook, and a spokesperson did not promptly reply to TechCrunch’s inquiry for comments. 

Independent cybersecurity reporter Brian Krebs mentioned in May that a security researcher from cyber firm GitGuardian alerted him to a multitude of exposed passwords located in a publicly available GitHub repository that had been uploaded by an employee of a CISA contractor.

As per Krebs, the researcher attempted to inform the contractor but received no response. Only after Krebs reached out to CISA did the agency take down the repository and revoke and replace all the compromised credentials to avert any potential misuse in the future.

CISA asserted that no customer or mission-related data was compromised in the incident and expressed gratitude to the researcher and journalist for their assistance. The agency admitted that its protocols for enabling security researchers to notify CISA of possible incidents “were not clearly defined,” and it has implemented changes to facilitate quicker communication for researchers wanting to contact the agency.

CISA has been without a permanent director since the commencement of President Donald Trump’s second term in January 2025. The agency has also experienced cutbacks, furloughs, and layoffs impacting around one-third of its workforce since Trump assumed office. 

Meta has discontinued a contentious feature that enabled users to alter images from public Instagram profiles using AI. The feature, introduced earlier this week alongside a suite of other AI functionalities, “didn’t hit the target” and is now off the table, as stated by the company.

This week, Meta unveiled Muse Image, a fresh AI image generator developed by Meta Superintelligence Labs, its specialized AI division. Meta highlighted one capability that allowed users to create images by @-mentioning public Instagram accounts they wished to cite. This feature, which wasn’t crafted to notify users if their pictures were utilized in this manner, sparked immediate criticism.

TechCrunch produced its own tutorial on how to turn off the feature.

Now Meta has changed its stance. The company released a blog post on Friday indicating it was retracting the feature. Puck News founding partner Dylan Byers was the first to report on the company’s decision.

“Our goal was to offer a beneficial creative resource and to empower individuals to decide if their public content could be referenced in this fashion,” the company stated on its blog. “We’ve received the feedback that this feature missed the mark, so it’s no longer available.”

TechCrunch contacted Meta for additional details and will refresh this article if they provide a response.

Since its adoption by social media networks, AI has often been misused — particularly to create explicit images of female celebrities. Platforms have tried to address this issue, although the measures taken have frequently been inadequate.

Concerning Meta’s recently removed feature, it appears quite clear that it could have been exploited. Indeed, Byers points out that the choice to eliminate the feature was made “in light of concerns from users and talent agencies, including CAA.”

In March, Bluesky’s former CEO, Jay Graber, stepped down to take on the role of chief innovation officer. He was promptly succeeded by Toni Schneider, the founding CEO of Automattic, the firm behind WordPress and Tumblr.

After serving as interim CEO for the last four months, Schneider is officially removing the “interim” title and becoming the permanent chief executive.

“I’ve been in my interim CEO position at Bluesky for four months, and it’s time for an update,” Schneider stated on his personal blog. “Most importantly, as of today, the interim designation has been eliminated. I am fully committed to Bluesky’s mission and responsibilities as its official CEO.”

In his post, Schneider mentioned that one of his initial goals is to “establish smaller spaces and more private communities,” which he believes will “catalyze the next phase of growth and innovation.”

Both Automattic and True Ventures, the venture capital firm where Schneider is a partner, are notable investors in Bluesky.

Originally a spin-off from Twitter, Bluesky has become a refuge for individuals seeking to escape the alterations Elon Musk implemented following his acquisition of the platform in 2022 (the platform was later renamed X and is currently a subsidiary of Musk’s combined rocket and AI venture, SpaceXAI).

Under Graber, the platform expanded to 43 million users, while the AT Protocol, the foundational technology facilitating Bluesky and other applications to share the same social network, underwent considerable enhancements.

Recently, however, the platform has faced challenges in maintaining or increasing its user base. Some have raised concerns about its viability — noting marked declines in both user engagement and the community as a whole. Bluesky experienced a significant surge in user numbers following Donald Trump’s re-election (when Elon Musk was heavily involved in politics), but the site seems to have experienced a downturn since that time.

In summary: Schneider has some significant challenges ahead. He appears ready to take them on, though. “We are just at the start of this narrative,” he remarked on Friday.

Prime Intellect, a burgeoning startup offering computational power and tailored software tools for businesses to create AI agents, has successfully secured a $130 million Series A funding round, achieving a valuation of $1 billion.

This substantial funding round was spearheaded by Radical Ventures, with contributions from Nvidia Ventures, Intel Capital, Dell Technologies Capital, Iconiq, and a variety of angel investors who are founders of prominent companies, such as Aravind Srinivas (Perplexity), Aaron Levie (Box), Winston Weinberg (Harvey), Jeff Wang (Cognition), and Brendan Foody (Mercor).

Established in 2024, Prime Intellect aims to empower organizations to train their own agentic systems independently of advanced AI labs. Although this objective would have seemed daunting a few years back, the advent of reinforcement learning methods, which reward successful completion of tasks while penalizing mistakes, enables businesses to act as their “own AI lab” by honing models for particular business functions.

Despite the possibility of bypassing centralized AI laboratories, the foundational infrastructure remains exceedingly intricate, leaving many organizations without the required expertise to piece together a production-ready system.

This is precisely where Prime Intellect steps in.

The startup has created what it dubs a “full stack” for AI agent development, encompassing compute access, a reinforcement learning framework, and assessment tools.

Prime Intellect’s platform operates as a marketplace, allowing clients to select and utilize only the necessary tools they require without being bound to a comprehensive system.

“They’ve integrated this in such a manner that they’re pioneering at the cutting edge in a cost-effective way,” remarked David Katz, a partner at Radical Ventures. He noted that while competitors may provide isolated components, Prime Intellect offers the functionality of a premier AI lab as a “one-stop shop” for development.

The startup’s methodology has garnered customers like Ramp, Zapier, and Flapping Airplanes, who compensate the startup for a hosted version of its offerings. This swift adoption has catapulted the company to an annual revenue run rate of $100 million.

This expansion is driven by concrete results. For instance, Ramp utilized Prime Intellect to create an agent that assisted the fintech in deriving answers from spreadsheets. “The outcome surpassed the frontier models in accuracy while operating at higher speeds and a fraction of the cost,” stated Ramp’s co-founder and co-CEO Karim Atiyeh.

Another pivotal element fueling Prime Intellect’s growth is the newfound awareness among companies that building upon leading-edge labs entails several risks.

Businesses are increasingly hesitant to share their proprietary information with OpenAI and Anthropic due to potential data control loss. They also express concerns about relying on models that can be abruptly disabled, as demonstrated with Anthropic’s Fable last month.

“How can I be certain that I’m not partnering with a company aiming to replace me and generalize my work,” Katz commented. “All of these factors are prompting people to consider, ‘How do I secure my own enterprise intelligence and mitigate these risks’.”

Co-founder and CEO Vincent Weisser of Prime Intellect believes that enterprises are eager to shift away from proprietary frontier models, and his company supplies the necessary infrastructure to facilitate this transition.

“It shouldn’t merely be a handful of tech enthusiasts in a glass tower in San Francisco with the ability to train AI models,” he expressed to TechCrunch. “It should encompass every enterprise, every nation state.”

U.S. insurance firm AssuranceAmerica has announced a data breach involving the personal details and driver’s license numbers of 6.9 million individuals, marking it as the most significant known leak of Americans’ driver’s license data this year.

Established in 1998, AssuranceAmerica offers car and rental insurance to clients in over a dozen U.S. states. Being a major insurance entity, the company manages extensive data concerning potential insurance clients and vehicle operators, encompassing their personal details and data regarding their state-issued driver’s licenses. In the possession of a malicious individual, a driver’s license number could facilitate fraud and impersonation.

In a data breach notification dispatched to clients and observed by TechCrunch, AssuranceAmerica revealed that it detected unauthorized access in its computer systems on March 17. The company completed its investigation on June 15, discovering that the intruders had extracted clients’ names, contact details, and driver’s license numbers. 

The breach notification stated that the hackers also accessed data concerning clients’ auto insurance policies and accounts, as well as information about their drivers and vehicles, and particulars about client claims. 

The company did not elaborate on what other types of personal data were compromised.

AssuranceAmerica did not disclose the exact reason for the breach but mentioned that the hackers “targeted one of the Company’s employees” and that the company subsequently “disabled compromised credentials.” It remains unclear how those credentials were acquired, but past incidents involving the theft of employee credentials have been associated with password-stealing malware or the use of compromised software.

TechCrunch sent inquiries regarding the situation to AssuranceAmerica CEO Joe Skruck and founder Guy Millner, including whether the company had engaged with the hackers or provided a ransom. Neither answered.

According to a data breach report filed with the Indiana attorney general’s office, AssuranceAmerica noted the breach as affecting 6.99 million individuals, with notification letters scheduled for distribution on July 10.

An alternate copy of AssuranceAmerica’s data breach notification, provided by the Maine attorney general’s office at TechCrunch’s request, also indicates the number of individuals affected is 6.99 million. (Maine’s data breach portal is currently offline and under review following a fraudulent breach disclosure published on its website last month.)

The occurrence at AssuranceAmerica coincides with a series of data breaches impacting driver’s licenses and other identity documents in recent months. In June, the Texas state government reported that hackers compromised data related to at least 3 million driver’s licenses and passport numbers during a breach affecting the state’s parks and wildlife division.

TechCrunch has previously covered numerous security flaws that collectively compromised millions of government-issued identity documents, including incidents involving a hotel check-in system, a money transfer application, a prison payphone provider, and a U.K. visa service. These data leaks arise as websites and apps increasingly require users to provide their identity documents to verify their legal age for access, amid a global initiative by governments to implement age-verification regulations.

As both established and emerging companies race to leverage AI, a number of AI startups report that their revenue is not just increasing but also accelerating rapidly, achieving milestones in shorter time frames.

The startups listed below have exhibited a trend of such flywheel growth. It’s important to note that the metrics these companies reference vary, even when they mention “ARR.” Some may define it as annualized recurring revenue (ARR) or contracted revenue from a paying client that has yet to be billed. Others may refer to annualized run-rate revenue, which estimates annual earnings by extrapolating from the revenue of the most recent month over 12 months. Additionally, some refer to “committed ARR,” which includes contracts signed by customers who are yet to be onboarded. In Gusto’s case, it reported its actual trailing revenue for the last 12 months.

Nonetheless, each of these startups, presented in reverse order based on when their ARR growth became known, indicates that their revenue growth is picking up speed, however that may be defined. There are certainly many more rapidly growing AI startups than those mentioned here, but we are focusing this list on companies achieving revenue milestones at increasingly swift rates.

Mercor: On Monday, co-founder and CEO Brendan Foody announced that the company has surpassed $2 billion in gross annualized revenue as of June — a mere four months after reaching the $1 billion benchmark. The firm, which employs domain specialists to develop and enhance AI models, stated that it attained a $500 million run rate in September.

Anthropic: This model maker has recently seen its revenue grow at such an unprecedented pace that it has captivated the entire AI industry. In late May, Anthropic revealed that it exceeded a $47 billion revenue run rate, a noteworthy achievement that came less than two months after announcing its revenue run rate had surpassed $30 billion. The company stated it reached a $9 billion revenue run rate by late 2025, up from a previously reported $4 billion in July 2025.

Sierra: After achieving its first $100 million in ARR within seven quarters, Sierra — which creates AI agents for enterprise customer service — announced in late May that it took just two additional quarters to add another $100 million, according to co-founder and CEO Bret Taylor.

Glean: In May, Glean declared that it surpassed $300 million in ARR. While it required the seven-year-old enterprise AI startup nine months to increase its ARR from $100 million to $200 million, the company claims it only took six months to boost that figure from $200 million to $300 million.

Gusto: The 14-year-old HR tech firm revealed in May that its revenue has accelerated every quarter for the past five. Valued at $9.3 billion in early 2022, the company also disclosed that it has exceeded $1 billion in trailing 12-month revenue. Gusto’s revenue growth demonstrates that AI-native companies aren’t the only ones experiencing significant top-line growth through the use of this technology.

Clio: This 18-year-old provider of legal practice management software saw a dramatic increase in revenue after integrating AI into its services in 2023. The company reached $200 million in ARR by mid-2024, doubled that amount by the end of last year, and recently announced that its ARR has hit $500 million.