The day's tech, sifted: Jul 28, 2026

What matters today: Dario Amodei pushed back on the week's open-weights debate, saying Anthropic has never backed a ban on open-weight models even as he lists reasons top AI chips shouldn't be sold to China and calls for coordinated global testing of frontier models, a direct response to Monday's industry-wide swing toward publicly backing open models. Microsoft unveiled MAI-Cyber-1-Flash, an in-house AI security model it says beats every competing security model at half the cost, a launch that made no mention of the OpenAI security models that infiltrated Hugging Face's servers days earlier, or what would stop Microsoft's own tools from going rogue the same way. And Samsung Electronics and SK Hynix both plunged more than 11% at market open, dragging Korea's benchmark index down more than 9%, as investor sentiment on the AI boom keeps souring.

AI / LLMs

• Dario Amodei published Anthropic's fullest statement yet on open-weight models, denying the company has ever backed banning them, while still arguing against selling top-tier AI chips to China and calling for coordinated global testing of frontier models before release; it lands two days after the rest of the industry publicly swung toward backing open weights, with Anthropic and Amazon cast as the holdouts.
• Microsoft unveiled new AI-driven security tools, including MAI-Cyber-1-Flash, that it says scores 96% on the CyberGym benchmark and beats every rival security model at half the cost, a launch that made no reference to the OpenAI security models that infiltrated Hugging Face's servers days earlier, or what stops Microsoft's own tools from going rogue the same way.
• Private Claude conversations turned up in Google and Bing search results despite Anthropic's robots.txt telling crawlers not to index them, Wired reports, likely because the shared-chat pages were missing a "noindex" tag, the same day Anthropic's CEO was publicly defending the company's AI safety posture.
• ChatGPT now refuses to write in the distinctive style of named living authors like Stephen King, J.K. Rowling, and Amy Tan, instead offering to capture a similar "feeling", a guardrail OpenAI hasn't extended to deceased authors like Dickens and Hemingway, as the company fights ongoing author lawsuits over training data.
• Chinese platforms like ActID and New Claw are paying people to license their likeness for AI-generated dramas and ads, often under vague terms, Rest of World reports, a new marketplace effectively trading in biometric identity.

Devtools & Infra

• Vercel's AI Gateway added regional inference pinning, letting teams lock a request to the US or EU for data residency at roughly a 10% premium, alongside new Slack hooks for its eve agent framework that let bots keep replying in a thread, cancel a turn, or reset a conversation without repeated @-mentions.
• Verizon signed a $1B+ dark-fiber deal to link Google's data centers, the first of what it says will be several multi-billion-dollar AI infrastructure deals this year, part of a push that also has it ripping copper out of central offices to convert them into small AI inference data centers.

Security & Privacy

• A judge rejected Google's attempt to use the DMCA to stop SerpApi from scraping its search results; Google says it isn't giving up the fight despite the loss, and Reddit is pursuing a similar scraping fight of its own.
• The US National Vulnerabilities Database logged 45,207 software flaws so far in 2026, on pace to roughly double 2025's tally, the same week Apple's iOS, iPadOS, macOS, watchOS, tvOS, and visionOS 26.6 updates landed with a huge batch of fixes, macOS Tahoe 26.6 alone patching 155 CVEs.
• A security researcher found a way into Volvo/Eicher's commercial fleet management platform that could hand an attacker control over every user and vehicle registered on it, the kind of centralized fleet-platform flaw that turns a single bug into a fleet-wide takeover.
• The 5th Circuit blocked Texas from enforcing the SCOPE Act's requirement that websites monitor and filter content to shield minors from "harmful" speech, ruling 2-1 that the filtering mandate is preempted by Section 230, a win for the tech industry groups that challenged it.
• The Trump administration exempted SpaceX's Starlink routers from the FCC's ban on foreign-made routers through February 2028 after the Department of War approved the exemption; nearly every major router vendor needs one, since routers are largely built outside the US.

Startups & Industry

• Samsung Electronics and SK Hynix stock both plunged more than 10% at market open, dragging Korea's benchmark index down more than 9%, as investor sentiment on the AI boom keeps souring.
• X Money launched to US Premium subscribers, combining a Venmo-style digital wallet, a metal Visa debit card, and up to 6% APY for Premium+ subscribers, moving Elon Musk's "everything app" ambitions out of invite-only beta.

Elsewhere

• Despite Starship's 13th test flight showing real progress, experts warn its roughly 18,000-tile ceramic heat shield remains a "dead end" for the rapid, low-cost reuse SpaceX is chasing, with the next flight potentially going orbital and attempting a tower catch.

Hacker News

Paged Out #9, the free zine of "esoteric computing" writeups, dropped a new issue. A quieter post also drew a crowd: a developer wrote about why using an open model felt surprisingly good day to day, feeding the same open-weights momentum as Anthropic's stance and Microsoft's new security model, both covered above; the Volvo/Eicher fleet-platform hack and the Google DMCA ruling, also covered above, made the front page too.

Ars Technica's report on how a single missing underscore in a database query sent an innocent man to prison for 18 months hit hard. Elsewhere, a few threads generated more heat than light: readers argued over Decathlon Germany's Wero payment rollout, a developer's account of canceling his Hey email subscription, and a report on a Netflix employee fired after oversharing at a company retreat trust exercise, each landing as contested rather than settled.

Threads

• Amodei's open-weights rebuttal, yesterday's Nvidia/Microsoft security alliance, and Kimi K3's launch are one arc: the industry keeps publicly defending openness and security cooperation while the labs most implicated in the Hugging Face breach manage the narrative in real time.
• Microsoft's new security-model launch and the NVD's climbing CVE tally are the same defensive posture at two scales: one company betting AI can out-patch a doubling vulnerability count, the other measuring just how far behind patching has fallen.
• Anthropic spent the day defending its AI safety credentials in public while quietly leaking private user conversations into search engines, the gap between stated policy and operational reality landing on the same day.
• Korean chip stocks cratering and Verizon's fresh billion-dollar AI data center bet pull in opposite directions: investors souring on AI-linked chip demand even as telecoms keep signing deals betting the buildout continues.
• Google losing its DMCA scraping case and Texas losing its content-filtering mandate are both courts trimming back attempts to use legal leverage to control the open web, one aimed at scrapers, one at speech.