2026-10-25 Evening edition
Evening edition — Research Report

AI News Daily 2026-10-25

Date
2026-10-25
Edition
Evening edition
Audience
Executives, decision makers and business leads
Format
Detailed research report
Executive Summary
  1. Anthropic has announced plans, subject to shareholder approval, to contribute 5% of its equity to the newly created Anthropic Public Benefit Fund, with a mechanism for the board to add up to 2% more each time company value rises by a set threshold.
  2. NVIDIA has agreed with six supply-chain partners — Amkor, IBIDEN, King Yuan Electronics, Samsung Electro-Mechanics, SPIL, and Unimicron — to substantially expand capacity for substrates, advanced packaging, optical engines, and semiconductor inspection through 2026–2028.
  3. Anthropic has started a long-term initiative, the "Anthropic Cyber Mission," to support organizations defending critical infrastructure and software against AI-enabled cyberattacks.
  4. OpenAI has added a "Locked Chats" feature to ChatGPT that lets users hide and protect specific conversations behind a PIN or device authentication, rolling out across Free, Go, Plus, and Pro tiers.
  5. OpenAI has retired the GPT-5.5 model for ChatGPT-authenticated Codex users, directing them to the successor model, GPT-6.1 Sol.

01 Anthropic plans to contribute equity to a new Public Benefit Fund

Published: 2026-10-18 09:00

Facts

Anthropic has announced a plan, contingent on shareholder approval, to contribute 5% of its equity to a newly established entity called the Anthropic Public Benefit Fund. The company is also introducing a mechanism that would let its board of directors add, at its discretion, up to a further 2% to the fund each time Anthropic’s company value increases by 50% from a defined baseline.

Background

The announcement comes as Anthropic has grown rapidly as an AI company, and as the industry more broadly faces questions about how fast-growing AI firms should formalize commitments to public benefit alongside commercial growth.

Implications

As a structural precedent for institutionalizing profit allocation and public-benefit commitments at a major AI company, this move could influence how other AI companies approach their own capital and governance policies going forward.

Source: Anthropic — "Public Benefit Fund"

02 NVIDIA locks in capacity-expansion deals with six supply-chain partners

Published: 2026-10-21

Facts

NVIDIA announced partnerships with six companies — Amkor, IBIDEN, King Yuan Electronics, Samsung Electro-Mechanics, SPIL, and Unimicron — to significantly expand production capacity for substrates, advanced packaging, optical engines, and semiconductor inspection. The expansion is planned to take place from 2026 through 2028.

Background

Substrates, advanced packaging, and inspection processes are widely regarded as bottlenecks in AI infrastructure production, directly affecting how much AI compute hardware can be manufactured and shipped.

Implications

This multi-year capacity build-out is a significant supply-chain investment that will influence the stability of GPU supply and related cost trends over the next several years.

Source: NVIDIA — "NVIDIA and Supply Chain Partners Expand Capacity to Accelerate AI Infrastructure Buildout"

03 Anthropic launches the "Anthropic Cyber Mission" for cyber defenders

Published: 2026-10-08 09:00

Facts

Anthropic announced a long-term initiative called the "Anthropic Cyber Mission," aimed at supporting organizations that defend critical infrastructure and software. The company said it will provide tools, research, and other resources to help strengthen the defensive side of cybersecurity.

Background

The initiative comes amid a rise in AI-enabled cyberattacks, a trend that has pushed AI developers themselves to consider what role they should play in supporting defenders rather than only addressing offensive misuse after the fact.

Implications

An AI company stepping directly into defender support provides a reference point that other organizations may draw on when shaping their own security strategies in response to AI-related threats.

Source: Anthropic — "Anthropic Cyber Mission"

04 ChatGPT introduces PIN-protected "Locked Chats"

Published: 2026-10-20

Facts

OpenAI added a "Locked Chats" feature to ChatGPT that allows users to hide specific conversations from normal chat history and search, protecting them with either a 4-digit account PIN or device-level authentication. The feature is being rolled out in stages across the Free, Go, Plus, and Pro tiers.

Background

The update responds to a need among ChatGPT users — both individuals and business users — to manage access to conversations that contain sensitive information.

Implications

By giving users more granular control over which conversations are visible and accessible, the feature expands the options available for managing sensitive information when using a cloud-based AI assistant.

Source: OpenAI — ChatGPT release notes

05 OpenAI retires GPT-5.5 in Codex

Published: 2026-10-14

Facts

OpenAI retired the GPT-5.5 model for Codex users who are signed in through ChatGPT, and is directing those users to switch to the successor model, GPT-6.1 Sol.

Background

This change is part of a broader, fast-moving pattern of model-generation turnover at OpenAI, where models used in developer tools like Codex are periodically superseded and withdrawn.

Implications

Development teams relying on Codex need to regularly check OpenAI’s model deprecation schedule and plan migrations in advance to avoid disruption when an underlying model is retired.

Source: OpenAI — Codex changelog

— Editor’s note

This evening edition is a recap: today’s monitoring window did not surface a new story that met this report’s sourcing bar, so the edition instead gathers notable items from the past few weeks that are each backed by a confirmed Tier 1 official source.

Read together, the five items point to a few consistent patterns. OpenAI continues to ship frequent, incremental updates to ChatGPT and Codex — this time a privacy-oriented "Locked Chats" feature and the retirement of an older Codex model in favor of GPT-6.1 Sol — reflecting a steady cadence of small changes rather than infrequent large releases. Anthropic, meanwhile, is reinforcing long-term commitments on two fronts at once: formalizing how company growth translates into public-benefit funding through the new Anthropic Public Benefit Fund, and taking a proactive role in cybersecurity defense through the Anthropic Cyber Mission. And NVIDIA’s multi-year capacity agreements with six supply-chain partners underline that the physical bottlenecks in AI infrastructure — substrates, advanced packaging, and inspection — remain a key constraint that hardware suppliers are actively working to resolve through 2028.