Top AI News – October 19, 2026

Published today, sourced from stories within the last 24 hours.

🔥 Top Stories

1. Anthropic commits 5% of shares to new Public Benefit Fund ahead of planned November IPO

Anthropic announced it will commit 5% of its shares (subject to stockholder approval) to a new wholly-owned LLC, the Anthropic Public Benefit Fund. Its board may add another 2% for every 50% increase in company value over a set baseline. Combined with earlier founder and employee commitments, roughly 20% of shares are now dedicated to public-benefit purposes. The Wall Street Journal separately profiled the company’s opaque five-member Long-Term Benefit Trust, which has already picked four of seven board directors and could replace the board majority — a first for a pre-IPO tech company.

Sources: Anthropic · AI Weekly / WSJ

2. Musk responds with pledge: one SpaceX share for every US kid in 2027

Hours after Anthropic’s equity announcement, Elon Musk said on X he will give one of his SpaceX shares to “every kid in America” in 2027 — a giveaway worth roughly $10.5B across ~72M American children — saying he prefers distributing shares “directly to the public.” No eligibility criteria or distribution mechanics were disclosed.

Source: AI Weekly

3. OpenCode confirms distillation attack hit 1.4% of Console users; rotates API keys with Google, OpenAI and Anthropic

OpenCode confirmed a large-scale AI distillation attack reached 1.4% of its Console users — limited to those who had connected external AI providers. Leaked user-stored API keys for OpenAI, Anthropic and Google potentially affected; the company is revoking keys, urging password changes, and working with all three labs on remediation as the investigation continues.

Sources: OpenCode on X · AI Weekly

4. Tenstorrent raises $1.5B Series E at $6.2B valuation

Jim Keller’s chip firm closed a $1.5B Series E led by Trace Capital and Innovation X, more than doubling its $2.6B December 2024 valuation. Backers include Fidelity, Morgan Stanley, ARK Invest, T. Rowe Price, SanDisk and Tether Investments. Proceeds will scale mass production of its RISC-V-based Galaxy servers.

Source: AI Weekly

5. Microsoft pulls out of Nscale’s planned Essex AI data center

Microsoft withdrew from Nscale’s planned Essex site, which had been positioned as one of the UK’s largest AI-computing projects. Nscale says it still intends to build, but loses a major customer and must find another route to bring the capacity online.

Source: BBC

6. DeepSeek opens Search & Fetch API to developers; ships V4.1 Flash

DeepSeek opened the web-search system behind its own app to outside developers at $0.75 per 1,000 successful searches — far below incumbents like Brave’s $5/1,000. It also released DeepSeek V4.1 Flash 1017, a faster agent/coding model with open weights, via its app and API. Reuters notes the cheap Chinese model is catching up with OpenAI and Anthropic “on their home turf.”

Sources: X · Reuters

7. Anthropic acquires Opaque Systems’ confidential-AI stack

In a licensing-and-acquihire deal, Anthropic picked up Opaque Systems’ CEO Aaron Fulkerson, CTO Rishabh Poddar and several engineers, gaining hardware-rooted confidential computing, cryptographic attestation and “TRACE” agent-governance tooling as it pushes enterprise, sovereign and on-prem deployments. Opaque (a UC Berkeley RISELab spin-out) had customers including ServiceNow and Accenture.

Source: AI Weekly

📈 Markets

  • Chip stocks rebound Monday: US chip stocks recovered some ground after their steepest weekly selloff in months, as lower Treasury yields and easing crude prices supported the battered AI trade. Nasdaq-100 futures were up ~1.3% and S&P 500 futures ~0.8% ahead of an earnings week that will re-test the AI trade. Asian markets advanced despite a roughly $2 trillion chip rout in the region. (Reuters / Bloomberg)
  • AI financing concerns linger: Analysts say the reversal is forcing investors to separate demand for compute from the financial returns of supplying it; last week CoreWeave bond yields neared 15%.
  • NextDC halts trading as a Melbourne AI data center lost power. (Kalkine)
  • SemiAnalysis forecasts a 50GW data-center year in 2027 — matching the cumulative pre-2024 global market in a single year — with ~$1T in binding lease commitments already signed by Meta, AWS, Microsoft, Google, Nvidia, AMD, Anthropic and OpenAI.

🚀 Products, Models & Launches

  • Sam Altman confirms OpenAI humanoid robot, targeting a 2027 demo.
  • GPT-6 Astra Pro parallel agents: OpenAI upgraded ChatGPT’s Astra Pro with faster responses and parallel multi-agent task execution for Pro users.
  • OpenAI GPT-Rosalind Academic Plan: eligible US university labs can apply for up to 30 free Business Standard seats for 12 months for life-sciences work (drug design, genomics).
  • ChatGPT voice expansion: 13 new voices and broader multilingual speech with GPT-Live-1.1.
  • doubleAI (Nvidia-backed) launched WarpSpeed, a GPU-kernel code-execution model, alongside a $220M raise and purchase of 1,000 Nvidia B300 chips.
  • Claude cracks decades-old crypto mystery: Anthropic says an internal model traced the unexplained seeds behind NIST elliptic curves to the lrand48 PRNG seeded with Unix timestamps — solving a 30-year-old puzzle in under half an hour.

🤝 Partnerships & Deals

  • Foxconn to buy 50% of Mitsubishi Electric Mobility for JPY 115B (~$727M), bundling EV powertrains and autonomous-driving software with Foxconn’s manufacturing scale.
  • SoftBank in early talks to buy Allied Universal (G4S owner) for up to ~$18B excluding debt, extending its AI/robotics security push.
  • ByteDance reportedly discussing a custom data-center chip with MediaTek, potentially adding a supplier alongside Qualcomm, pending US regulatory approval.
  • Goldman Sachs poaches Google’s Evan Kotsovinos as partner and head of asset & wealth management engineering.

🏛️ Policy & Regulation Watch

  • China’s draft export controls on AI technologies and chips (unveiled late last week) continue to ripple through markets as the week’s earnings test the AI trade.
  • WSJ flags Anthropic’s non-shareholder Long-Term Benefit Trust as a novel governance test ahead of its planned November IPO — no tech company has gone public with such a structure.

Disclaimer: Published today, sourced from stories within the last 24 hours (October 18–19, 2026). Details may evolve as stories develop.

Leave a Reply

Discover more from AI Master's Blog

Subscribe now to keep reading and get access to the full archive.

Continue reading