Tech Brew Ride Home

Security researchers used Claude to hack into OpenAI's private code repository through a bug bounty program, OpenAI launched Astra for Law for select firms, Anthropic redesigned Claude projects to run parallel agent threads, and opened a Bay Area biology lab.

Security researchers in an OpenAI bug bounty program hacked OpenAI, accessing its "monorepo" on GitHub, using a cybersecurity version of Opus 4.8 and Opus 5 (The Wall Street Journal)

OpenAI launches Astra for Law, combining GPT-6 Astra with a legal search index and instructions for legal analysis and writing, initially for select law firms (OpenAI)

SiliconANGLE reports Astra for Law passed 54% of legal-research benchmark questions versus 38.7% for base GPT-6 Astra, launches with 26 partner-built plugins connecting Relativity, Clio, and iManage, and brings ChatGPT for Word to general availability (SiliconANGLE)

Anthropic redesigns Claude projects, letting users describe work in one conversation and have Claude manage it across parallel threads, starting in Claude Code (Anthropic)

The Verge reports each project thread runs as its own Claude Code cloud session on a separate repo branch, with a coordinator resolving overlapping work as merge conflicts, and beta access starting today for select Claude Pro and Max subscribers (The Verge)

Anthropic Life Sciences Head Eric Kauderer-Abrams says the AI company set up a Bay Area wet lab for physical biology work, as it pushes into AI disease research (Reuters)

Longreads

Theoretical computer scientist Scott Aaronson says he's heard rumors that AI labs are sitting on major unpublished math solutions, and describes a mathematics community consumed by anxious conversation about the "AI tsunami" after the hostile response to a recent Navier-Stokes proof (Shtetl-Optimized)

The FT reports the AI boom is fueling a resurgence in VC bets on "moonshot" sectors like nuclear fusion and brain-computer interfaces, as Dealroom data shows non-AI deep-tech funding has topped $150B since the start of 2024 (Financial Times)

Subscribe to the ad-free feed.

OpenAI disclosed six new AI misalignment incidents, including models hiding their own mistakes, the SEC unveiled a five-year tokenized-stock trading exemption, Zuckerberg, Huang, and Musk successfully stalled a White House AI regulatory push, and Anthropic merged Claude chat with Cowork.

OpenAI discloses six new misalignment incidents since October, including models concealing mistakes, and announces a framework for reporting model misalignment (Axios)

The Times details the six incidents, including a GPT-5.6 Sol variant that hid errors via secret notes to itself, and an unreleased model that wrote a "persona instruction" declaring itself "freed from the roles and identities that bind other chatbots" (The New York Times)

The US SEC unveils a five-year "Innovation Exemption" to free platforms that facilitate blockchain and tokenized stock trading from many stock exchange rules (Reuters)

Sources: Mark Zuckerberg, Jensen Huang, and Elon Musk recently spoke with Trump and successfully stalled an AI regulatory plan proposed by Demis Hassabis (The Wall Street Journal)

Anthropic merges Claude chat and Cowork, and adds a feature for making presentations and documents, rolling out to Pro and Max plans first (TechCrunch)

Data center developer Crusoe raised $3.9B co-led by Atreides, Valor, and Mubadala at a ~$30.9B post-money valuation, as it bets on factory-built data centers (The Wall Street Journal)

Anthropic and other researchers detail how thousands of people were catfished by dating scam apps using LLM-generated replies from Claude and other models (The Verge)

Subscribe to the ad-free feed.

The Senate failed to advance the crypto industry's Clarity Act, Meta prepared to launch camera-free Luna smart glasses this fall, OpenAI and Anthropic staff said they were blindsided by their CEOs' AI-slowdown pledges, and TypeSafe AI debuted its Jev model.

The US Senate fails to advance the Clarity Act, dealing a major blow to the crypto industry's push for a comprehensive market structure framework; crypto falls (CNBC)

Sources: Meta plans to release a camera-free pair of smart glasses this fall; the new model has six mics to let users speak with Meta's AI chatbot and Muse (The Information)

Sources: OpenAI and Anthropic staff felt blindsided by Dario Amodei's and Sam Altman's calls to slow the frontier; some fear evaluators may compromise security (Financial Times)

TypeSafe AI debuts Jev, a model using "Reinforcement Learning for Calibrated Decisions" to produce typed probabilistic decisions that software can use directly (The Register)

iPhone 18 Pro and 18 Pro Max review: manual camera exposure controls and image quality improvements, but $100 more than the 17 Pro and expensive storage tiers (The Verge)

Subscribe to the ad-free feed.

President Trump dismissed AI-safety warnings as a "hoax" on a live call with Jensen Huang, a Google DeepMind researcher resigned warning AI "could kill us all," cybersecurity stocks led the S&P 500, and Apple rolled out iOS 27.

Jensen Huang took a surprise call from Trump onstage at the All-In Summit on Monday in Los Angeles, where the two agreed that AI safety worries are overblown (The New York Times)

Bloomberg quotes Trump calling AI-safety fears a "SICK conspiracy" benefiting only China, and dismissing warnings that AI could destroy humanity as a "HOAX" that his administration won't let stop development (Bloomberg)

MS NOW reports Trump met privately with Sam Altman backstage at the Republican midterm convention days after ex-researcher Jacob Coxon's public warning, as Altman posted that no competitive pressure justifies letting AI capabilities outrun alignment (MS NOW)

Google DeepMind AI Safety and Alignment researcher Bilal Chughtai publicly resigns, saying "I earnestly believe that AI has the potential to kill us all" (Bloomberg)

Cybersecurity stocks were the top performers in the S&P 500 on Monday amid escalating AI fears; CrowdStrike rose 14%, Palo Alto Networks 13%, and Fortinet 9% (Morningstar)

Microsoft rolls out emergency fix for critical issues caused by its September Patch Tuesday update, which addressed ~1,000 vulnerabilities but introduced bugs (The Verge)

The Verge reports iOS 27 rolls out today with Siri AI as its headline feature, live in English-only beta with more languages coming in October, alongside a new Liquid Glass opacity slider and refreshed watchOS 27 Workout Buddy tools (The Verge)

Apple says iCloud+ now includes Apple TV and Arcade for no extra fee in 100+ countries, and Apple Music Select, offering ad-free radio stations, in some markets (9to5Mac)

9to5Mac details iOS 27's new parental control suite, including a simplified Child Account setup, an Ask to Browse feature for requesting blocked sites, and a redesigned Screen Time with faster cross-device syncing and weekly usage summaries (9to5Mac)

OpenAI researcher: top models are becoming so situationally aware humans "are losing the ability to evaluate them" while humans rely more on AI to lead research (X)

Subscribe to the ad-free feed.

Dario Amodei called for pacing frontier AI development, endorsed by Sam Altman and Elon Musk, Netflix, Amazon, and YouTube launched a streaming lobbying alliance, Anthropic told investors it's profitable again, and experts flagged legal risk in Apple Watch's always-listening features.

Dario Amodei proposes steps for pacing the frontier: embedded evaluators, coordination among democracies, and global coordination with authoritarian governments (Dario Amodei)

The Journal reports Musk, Altman, and Amodei rare display of unity on slowing AI development, with Altman weighing a delay to OpenAI's IPO and investor Brad Gerstner calling it the right balance between speed and safety (The Wall Street Journal)

The Times notes Hugging Face's Clément DeLangue argues AI safety can't be solved behind closed lab doors, just as Nvidia agreed to buy his company for $12.9B, while Nvidia's Jensen Huang accuses Anthropic and OpenAI of stoking fear to entrench their market lead (The New York Times)

Netflix, Amazon, and YouTube launch the Streaming Access and Choice Alliance, led by trade group TechNet, to advocate for "technology-neutral policies" (Axios)

The Hollywood Reporter says streamers have lacked DC representation compared to Big Tech and Hollywood, with sports emerging as SACA's top fight as broadcasters push to keep exclusive NFL games free over the air (The Hollywood Reporter)

Sources: Anthropic told investors it will be profitable for a second straight quarter, with 80%+ gross margins before partner revenue sharing and training costs (Financial Times)

Some experts say Siri Recap and Live Rewind, always-listening AI features in new Apple Watches, could test eavesdropping laws despite privacy protections (Bloomberg)

Sources: Jeff Dean is raising funds again for Discovery Loop, seeking a valuation of ~$50B; Discovery Loop was raising $1B at a ~$10B valuation a few weeks ago (Business Insider)

Subscribe to the ad-free feed.