TECH
Tesla says driver manually overrode autopilot in fatal Texas crash
A Texas crash in which a Tesla Model 3 struck a home, killing a 76‑year‑old woman, has drawn attention to the company's driver‑assistance technology. Tesla, through statements by AI vice‑president Ashok Elluswamy and Elon Musk, said the driver manually overrode the system by flooring the accelerator to 73 mph, a detail The Verge noted as the pedal being pressed “all the way to 100%,” while TechCrunch highlighted Tesla’s rare public pushback. Federal investigators from NHTSA and local authorities are reviewing the vehicle data to determine whether Autopilot or Full Self‑Driving was active.
Meta launches $299 Meta Glasses, its own brand of smart eyewear
Meta announced the launch of Meta Glasses, a $299 line of display‑less smart glasses made with EssilorLuxottica and branded solely as Meta, moving away from Ray‑Ban and Oakley labels. The glasses include a camera, personal speakers, a customizable AI‑assistant button, eight‑hour battery life plus a charging case for 40 extra hours, and upcoming pedestrian navigation and expanded language translation. The rollout follows Snap’s recent release of its high‑priced Specs, underscoring increasing competition in the wearable market.
Anthropic launches Claude Tag, an always‑on AI assistant for Slack
Anthropic announced Claude Tag, a new “always‑on” version of its Claude AI that lives in Slack and can be summoned with @Claude to provide insights, assign tasks, and retain contextual memory across channels. In beta for Claude Enterprise and Claude Team customers, the tool lets administrators define which channels and tools the AI can access, and includes an ambient mode that proactively updates teams and follows up on forgotten threads. TechCrunch notes the launch fits a broader industry push for enterprise‑focused AI, with rivals such as Microsoft, Snowflake and Databricks building similar context‑aware assistants.
Meta to develop prediction‑market app “Arena,” says Zuckerberg
The New York Times reports that Zuckerberg has ordered Meta to build a Polymarket‑style prediction‑market app, internally dubbed “Arena,” which would start as a points‑based game without real money. TechCrunch notes that while the project is a top priority, it could later incorporate monetary betting and that legal scrutiny of prediction markets is rising, with cases involving Polymarket, Kalshi and investigations of public figures. Sources say the app would be separate from Meta’s other platforms but could be promoted through them.
Oracle cuts 21,000 jobs to fund debt‑driven AI expansion
According to Ars Technica, Oracle announced a 12.9% workforce reduction, cutting about 21,000 jobs as it ramps up AI‑focused data‑center investments. The cuts are part of a restructuring plan that coincides with a $45‑$50 billion capital raise, half of which will be financed by debt, pushing the company’s total liabilities above $120 billion. Ars also notes that bondholders have sued Oracle for allegedly concealing the need for debt to build its AI infrastructure, and that the layoffs aim to improve cash flow amid concerns over reliance on customers such as OpenAI.
Klue breach linked to 2022 credential exposes LastPass and other customers
Klue said a legacy OAuth token from a 2022 pilot was used by hackers to steal data from its corporate customers, including password manager LastPass. The intrusion was detected on June 12 and publicly disclosed on June 23, with the Icarus hacking group claiming responsibility and threatening to release the data. Klue announced a review of its credential management, vendor‑access controls and monitoring processes.
Menlo Ventures raises $3 billion fund boosted by Anthropic investment
Menlo Ventures announced a $3 billion fund, the largest in its 50‑year history, largely driven by its early $750 million bet on AI startup Anthropic, now valued at roughly $14 billion. The firm structured most of the capital through a special purpose vehicle and added $250 million of its own money, later creating a $100 million Anthropic‑linked startup fund that has grown to about $250 million and backed over 60 AI companies. The fundraising reflects a renewed VC appetite for AI after the post‑pandemic slowdown, as reported by TechCrunch.
Microsoft, Chevron plan massive gas‑powered data center in West Texas
Microsoft and Chevron announced a 2.67‑gigawatt natural‑gas power plant in West Texas to supply a dedicated Microsoft data center for AI and cloud workloads. The 20‑year power purchase agreement will make the project one of the largest co‑located gas‑power and data‑center developments in the U.S., though it raises concerns about Microsoft’s 2030 carbon‑neutral target. The plant, called Project Kilby, is projected to emit over 13 million tons of CO₂ and other pollutants, according to the Environmental Integrity Project.
Superhuman acquires GPTZero to add AI detection to email
TechCrunch reported that Superhuman, the email platform formed after Grammarly’s acquisition, bought AI‑detection startup GPTZero, though the purchase price was not disclosed. Business Insider noted GPTZero previously amassed more than 19 million users and about $30 million in annual recurring revenue, having raised $13.5 million and become profitable in 2024. Superhuman said it already had a detector and that “two AI detectors are better than one,” prompting the move to integrate GPTZero’s technology.
Meta pauses employee‑tracking AI program after internal data leak
Meta has halted its Model Capability Initiative, an AI system that records employees’ keystrokes and mouse movements, following a leak that exposed private conversations and performance data to the broader staff. Engadget highlighted that the program allowed the entire company to view the sensitive information, while Business Insider reported that Meta said no evidence of improper access yet, prompting a pause pending investigation. The move follows previous AI‑related security issues at Meta, including an unprompted AI action in March and a chatbot breach earlier this month.