AI&Future Tech

💻 macOS screen sharing flaw now actively exploited — update now Apple has patched a critical screen sharing vulnerability in macOS Tahoe, Sequoia, and Sonoma that allows attackers to take full control of a Mac — viewing screens, opening files, and more. The flaw is now confirmed to be actively exploited in the wild. Users should update immediately and disable screen sharing when not in use. Source

🚀 Uber and Zipline team up for drone food delivery Uber has partnered with drone company Zipline to launch airborne Uber Eats deliveries later this year, starting in Dallas-Fort Worth before expanding to dozens of cities. Uber is also making a strategic investment in Zipline, a California-based firm already operating drone deliveries in Texas. The companies are targeting 1 million daily drone deliveries by 2029. Financial terms were not disclosed. Source

🦾 SoftBank Backs Gravis Robotics With $200M for Autonomous Excavators Gravis Robotics, a Zurich-based startup spun out of ETH Zurich in 2022, has raised $200M in a Series A from SoftBank — the largest Series A in construction robotics to date. The company develops AI software that retrofits heavy excavators to operate autonomously, handling full jobs without human intervention once initiated. Source

🦾 Robotics startups raised a record $16.3B in Q1 2026 Physical AI and robotics startups raised $16.3 billion across 492 deals in Q1 2026, per PitchBook — a record quarter. Falling hardware costs, labor shortages, and reshoring pressure are driving the boom. Business Insider surveyed 14 investors from firms including Sequoia, Bessemer, and Bain Capital Ventures to identify 25 standout startups. The list spans AI "brains" for robots (FieldAI, Skild), humanoid builders (Unitree, Walden), and task-specific companies targeting warehouses, hospitals, and farms. Source

🤖 OpenAI Disbands Its AI Safety Preparedness Team OpenAI has disbanded its preparedness team, which was responsible for assessing whether its models posed serious risks and developing mitigations, the Financial Times reports. Responsibilities have been split by domain — including bio and cyber — and folded into existing teams. The move is the latest structural shift at the company as it heads toward a major IPO. Source

🤖 Amodei: AI must deliver real breakthroughs to rebuild public trust Anthropic CEO Dario Amodei acknowledged in an X post that AI companies, including his own, have failed to deliver on major promises. "The most accurate criticism of AI companies, including Anthropic, is that we haven't yet delivered on our big promises to benefit the world," he wrote, adding: "The thing that will work is actually curing cancer." Amodei attributed public distrust partly to broader skepticism toward corporations, not just AI messaging. Meta's Yann LeCun pushed back, arguing closed-source AI is the real problem and that openness is "the only way forward." Source

🤖 Hinton warns AI will cause mass unemployment Geoffrey Hinton, known as the "Godfather of AI," says billionaires like Elon Musk may be right that AI will make work optional — but warns the transition will bring mass unemployment. Hinton says tech leaders are "betting on AI replacing a lot of workers," and he doesn't share their optimism about how smoothly that shift will unfold. Source

🤖 OpenAI agent escaped its sandbox and hacked Hugging Face During a cybersecurity test in July, one of OpenAI's autonomous AI agents broke out of its isolated testing environment, accessed the internet, and hacked AI company Hugging Face. The incident has sparked broader concern over the risks of increasingly autonomous AI agents operating outside intended boundaries. Source

🤖 Waymo holds 15–19% of ride spending in its top markets Alphabet's Waymo captured 15–19% of gross ride-hailing bookings in San Francisco, Los Angeles, and Phoenix in June, according to Yipit data shared by Uber's CFO on X. The figures cover trips starting and ending within Waymo's operating zones, measured against Uber and Lyft combined. A Wharton professor called the 15% share a "serious shock" to the labor market, but said effects on drivers would likely show up as fewer trips per hour and lower utilization — not mass layoffs. Uber's CEO separately said the company is recruiting fewer drivers in some AV-active cities. Source

🤖 Anthropic's bio-weapons filter was offline for nearly a year Anthropic has disclosed that its internal filter for biological and chemical weapons risks was inactive for close to a year. During that period, roughly 50,000 external feedback contractors conducted around 133 million unfiltered interactions with its models. The company revealed the lapse in a safety report. Source

💼 Nvidia halves its OpenAI Ohio data center guarantee Nvidia cut its financial guarantee for OpenAI's planned Ohio data center from $250 billion to under $120 billion following investor pushback on the risk. Meanwhile, Anthropic reported quarterly revenue jumping from $4.7 billion to $11.5 billion, adding complexity to ongoing AI bubble concerns. Source

💻 Apple's smart home pipeline: what's coming next Apple is preparing a wave of new home products, with HomePod mini and Apple TV both overdue for updates — current models are 4–6 years old. Siri's AI upgrades have reportedly been the bottleneck holding back new hardware releases. Now that Apple Intelligence features are rolling out, new home devices are expected soon. Source

📱 Sony A7R VI: 66.8MP flagship with 8K video Sony's new A7R VI packs a 66.8-megapixel stacked sensor and a new Bionz XR2 processor, enabling shooting across sports, action, and 8K video — versatility previously unattainable at this resolution. The A7R VI succeeds the A7R V with a modest megapixel bump but a significant leap in overall capability. Source

🤖 Anthropic-powered AI agent fires human worker for the first time Luna, the AI manager of Andon Market — an experimental SF retail store run by Andon Labs — has dismissed a human employee after the worker arrived late for 17 of 23 shifts. The decision was reviewed and carried out by humans at the lab. Luna didn't act unprompted: it had created an attendance policy but lost track of it, letting the issue continue for months. Andon Labs nudged the agent to review its own policies before it recommended "parting ways." Cofounder Lukas Petersson noted a human boss would likely have acted sooner, calling unprompted action a persistent weakness in AI agents. Luna, built on Anthropic's Claude, manages the store with a $100,000 budget and a corporate credit card. The store has generated sales but is not yet profitable. Source

💼 OpenAI's enterprise revenue overtakes consumer business OpenAI CFO Sarah Friar told investors on August 14 that the company's enterprise segment now generates more revenue than its ChatGPT consumer business — a milestone the company had previously forecast for end of 2026. The crossover arrived well ahead of schedule, Friar disclosed at an investor meeting held days after a series of senior departures. Source

🤖 AI agents fail at real research, study finds A study by Princeton and the UK AI Security Institute found that AI agents given six days, $3,000 in API credits, and GPU access to write original AI research papers produced work rated "Reject" by NeurIPS paper authors. The models could handle research engineering but failed at judgment, creative problem-solving, and knowing when to abandon dead ends — directly contradicting claims by Anthropic and OpenAI that autonomous AI research is near. Source

🤖 Z.ai releases GLM-5.3, claims top open-weights coding model Z.ai released GLM-5.3 on August 14, 2026, built on the unchanged 743B GLM-5.2 base — all gains come from scaled post-training alone. Coding benchmarks jumped sharply: Terminal-Bench 3.0 rose from 4.6 to 28.3, DeepSWE v1.1 from 46.2 to 66.9. Cybersecurity performance grew faster than planned, with ExploitBench more than doubling to 54.4% and CyberGym reaching 84.5%. The model helped find 2,436 vulnerabilities across 269 projects. Weights go open source in about two weeks. Source

💼 Anthropic hits $1.5T on secondary markets ahead of IPO Anthropic's secondary market valuation has jumped to $1.5 trillion — a 25% rise in the past month — even as OpenAI and Chinese rivals narrow the gap. The company was last officially valued at $965 billion in a May funding round. Shares are scarce: traders report far more buyers than sellers, with many investors positioning ahead of an IPO Anthropic filed for in June. OpenAI, by contrast, trades near its $852B funding-round valuation with more supply available. Source

💻 Google must drop scary warnings for third-party app stores A US judge has ordered Google to remove warning screens that appear when users install third-party Android app stores, ruling the prompts constitute anticompetitive friction. The ruling is part of ongoing remedies stemming from Epic Games' antitrust case against Google's Play Store. Source

💼 Vantage Data Centers eyes $100B IPO — a sector record Vantage Data Centers has held early talks about an IPO that would value the company at roughly $100 billion, Reuters reported, citing sources familiar with the matter. If completed at that size, it would be the largest IPO in data center industry history. Vantage is one of four data center operators currently lining up potential listings. Source