Subject: Failed Companies Are Selling Old Slack Chats and Email Archives to Train AI
Source: Gizmodo
https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916
Startups that are shutting down are now selling off their company data, including emails and Slack messages, for as much as hundreds of thousands of dollars to help train AI models.Forbes reports that companies that specialize in winding down startups are helping founders squeeze out some last-minute cash by monetizing their internal communications.While large language models were initially trained on public internet data like books, news articles, Wikipedia, and Reddit threads, newer agentic AI models require more complex datasets that reflect how work actually gets done.That training often happens in so-called “reinforcement learning gyms” or RL gyms. These simulated environments are built using real-world company data and allow AI agents to practice completing workplace tasks like planning a birthday celebration for a coworker. This kind of training data has quickly become very lucrative. The Information reported last year that leaders at Anthropic discussed spending up to $1 billion on RL gyms….According to its website, SimpleClosure helps companies determine what data can be sold, assess its value, and process it to remove personally identifiable information….Still, some privacy advocates are raising concerns about what this means for workers whose data may be included in those datasets.
…
The Center for AI and Digital Policy also recently sent a letter to the Senate Commerce Committee urging the Federal Trade Commission to step up its oversight of AI-driven businesses.
Explore more on these topics:
Source: Gizmodo
https://gizmodo.com/sam-altmans-creepy-eyeball-scanning-company-gets-in-bed-with-zoom-and-tinder-2000748013
No one wants to talk to a bot, but how far are you willing to go to prove that you’re human? Sam Altman is banking on people being willing to surrender scans of their eyes in order to authenticate themselves, and he’s amassing some powerful sources to push more people to go along with the scheme. On Friday, both Tinder and Zoom announced partnerships with Altman’s World, the company behind the creepy, eyeball-scanning orb that is meant to prove users are human.
World has already been working with Tinder and ran a pilot of the verification process in Japan. It was apparently enough of a success that Tinder will roll out the authentication method globally. According to a press release, users will be required to undergo World’s verification method, which requires having their eyeballs scanned at a physical location with a proprietary device to prove they are human. Once they do so, they’ll get a badge on their profile to signal that they are a verified human. Tinder will also tempt people to partake by offering five free “boosts”—a feature that temporarily makes a person’s profile appear first for other users.
Basically, every version of World’s approach to verification requires people to go along with invasive biometric scans, and while that might be the best way to ensure a person is who they claim to be, it’s also a pretty major uphill battle to sell people on submitting themselves to scans like this. The power of World’s partners will certainly make things trickier. Zoom is still a big deal across many workplaces, and Tinder remains one of the largest dating platforms in the world. We’re one step from these types of scans being mandatory, with the potential penalty of getting cut off from essential services for failure to comply.
Source: Government Executive
https://www.govexec.com/pay-benefits/2026/04/senators-demand-opm-withdraw-plan-access-feds-medical-records/412961/
More than a dozen Democratic lawmakers warned that a little-scrutinized proposal to collect claims-level data related to the Federal Employees Health Benefits and Postal Service Health Benefits programs could violate federal law and doctor-client confidentiality.
A group of 16 Democratic senators on Monday called on the Office of Personnel Management to withdraw its plan to collect claims-level health data from federal workers and retirees, expressing “grave concern” that the measure would violate the Health Insurance Portability and Accountability Act and basic tenets of doctor-patient confidentiality.
Last December, OPM published an information collection request in the Federal Register that would require insurers who participate in the Federal Employee Health Benefits and Postal Service Health Benefits programs to provide monthly reports with identifiable health data on their enrollees, prompting unease from both health ethicists and health care providers alike. The notice would require the collection of medical visits, prescriptions and treatment data, and fails to task insurance carriers with redacting personally identifiable information.
“Such sweeping access to personal health information would violate the core principles of the Health Insurance Portability and Accountability Act, which was enacted to strictly regulate how protected health information can be disclosed to ensure that patient data is shared only for limited, clearly defined purposes,” they wrote. “Mass, centralized access to identifiable medical records absent individualized consent, clear necessity or narrowly tailored legal authority undermines those protections and lacks a valid statutory basis.”
…
Filed: Pay & Benefits
Subject: We Don’t Really Know How A.I. Works. That’s a Problem
Source: New York Times
https://www.bespacific.com/we-dont-really-know-how-ai-works-thats-a-problem/
The New York Times: “For us to trust it on certain subjects, researchers in the growing field of interpretability might need to learn how to open the black box of its brain… A.I. system is to ask the model to explain itself. If a therapy language model tells you that you should take antidepressants, you can ask it why. “You have mood swings,” it might respond. “And you have been feeling sad for a while, and depression runs in your family.” Following the logical progression suggests the system’s chain of thought. This is what we do when other people make decisions. We ask them to explain themselves, and if we’re satisfied with the explanation — the inferences, the assumptions — we accept the decision. But this won’t do for most medical models. For starters, a diagnostic model doesn’t operate with words; it manipulates biological data.
…
Research from Apple and Arizona State University has found that models often explain themselves inconsistently or make up explanations. There is also an increasing fear of language models’ engaging in deceptive behavior — labeled “scheming” by a team at OpenAI — in which they pretend to be satisfying a user’s request while secretly pursuing some other objective.
…
You can’t trust an A.I. to translate the motives of another A.I. when all A.I.s are suspect…”
Source: The Register
https://www.theregister.com/2026/04/22/google_unleashes_even_more_ai/
Google Cloud chief operating officer Francis deSouza has summed up his company’s security strategy du jour as follows: “You need to use AI to fight AI.” That also sums up all of the security services and products announced Wednesday at Google Cloud Next – and every other tech firm’s strategy at this point in 2026. Google’s version of this plan essentially boils down to deploying more AI agents to hunt for threats, plus more tools to secure this expanding AI agent fleet.
“It is very clear that we have moved from a human-led defense strategy, to a human-in-the-loop defense strategy, to an AI-led defense strategy that’s overseen by humans,” deSouza told reporters during a press conference ahead of Google’s annual shindig in Las Vegas, happening this week. “Our model for the future is an agentic fleet that does a lot of the routine cyber security work at a machine pace and then is overseen by humans.”
More about:
Filed: Security
Source: Bloomberg News
https://www.bespacific.com/anthropics-mythos-model-is-being-accessed-by-unauthorized-users/
Bloomberg (Gift Article): “A small group of unauthorized users have accessed Anthropic PBC’s new Mythos AI model, a technology that the company says is so powerful it can enable dangerous cyberattacks, according to a person familiar with the matter and documentation viewed by Bloomberg News. A handful of users in a private online forum gained access to Mythos on the same day that Anthropic first announced a plan to release the model to a limited number of companies for testing purposes, said the person, who asked not to be named for fear of reprisal.
Subject: Online Betting Is Fueling a Wave of Bankruptcies Among Young Americans
Source: Business Insider
https://www.businessinsider.com/online-sports-betting-gambling-bankruptcy-young-americans-bankruptcy-2026-4
- Online sports betting is leading some young Americans into massive amounts of debt.
- Consumer bankruptcy attorneys say they’ve seen a surge of young clients with gambling-related debt.
- “They’re betting hundreds of dollars per hour, and not really knowing it,” one attorney said.
The online sports betting boom is pushing some Gen Z and young millennial gamblers into bankruptcy.
Consumer bankruptcy attorneys told Business Insider that over the past year or so, they’ve seen a surge of young clients — mostly men in their 20s and 30s — running into financial trouble largely due to online gambling.
In many cases, those clients have racked up tens of thousands of dollars in credit card debt by placing bets on online sportsbooks, the lawyers said.
…
The Federal Reserve study found that states that have legalized mobile sports betting have seen credit delinquency rates spike, particularly among borrowers under 40.
…
“There are ways to turn credit cards into cash,” said Boltz. “Or it may just simply be that instead, they use their credit cards for everything else, and their paycheck and their debit cards for the gambling, so then you start putting all of your other expenses on your credit cards.”