AI discovery: 2026/01

Date

2026/01/31

Cover

19.01.2026 – 25.01.2026

Introducing OpenAI’s Education for Countries — New “OpenAI for Countries” pillar to help governments and university consortia embed AI tools/training into national education systems.
Our approach to age prediction — ChatGPT rolls out age prediction to automatically apply under‑18 safeguards (with an option to verify age instead).
Horizon 1000: Advancing AI for primary healthcare — New initiative aimed at scaling practical AI use in primary healthcare.
ServiceNow powers actionable enterprise AI with OpenAI — ServiceNow + OpenAI partnership framing enterprise workflows and deployment.
Stargate Community — OpenAI describes an initiative focused on expanding AI infrastructure and access.

Building a Discovery pipeline with prompts — Figma: new prompt tooling/flow to build “Discovery” pipelines (Jan 23 entry).
New features in Make — Figma: Make gets embeds for prototypes, image layers, and more ways to connect to live data (Jan 22 entry).
Adobe partners with artists to power the new era of media & entertainment with Firefly Foundry — Adobe “Foundry” positions Firefly models as tunable to brand/franchise IP for commercially-safe generation.
New AI‑powered video editing tools in Premiere + major motion design upgrades in After Effects — Adobe adds tighter Firefly-to-video workflows and new AI-assisted editing capabilities.

Unrolling the Codex agent loop — Engineering write-up on how Codex’s agent loop works and how to improve it.
Scaling PostgreSQL to power 800 million ChatGPT users — OpenAI engineering deep dive on scaling Postgres for very large workloads.
Codex is now integrated into JetBrains IDEs — JetBrains adds an in-IDE Codex agent option with multiple auth modes (JetBrains AI / ChatGPT / BYOK).

Amazon Bedrock AgentCore: Using browser extensions — AWS docs for creating AgentCore browser sessions with custom extensions (useful for automation/testing/agent browsing).

GitLab patches major security flaw (CVE-2026-0723) allowing 2FA bypass — Summary of a GitLab auth issue and recommended upgrades for self-managed instances.
Anthropic Git MCP server vulnerabilities (MCP/agent security) — Report on critical Git MCP server flaws with CVEs and patch details.
NVD: CVE-2025-68143 — Anthropic Git MCP server: unrestricted git_init issue (per reporting).
NVD: CVE-2025-68144 — Anthropic Git MCP server: argument injection in git_diff (per reporting).
NVD: CVE-2025-68145 — Anthropic Git MCP server: path validation bypass (per reporting).

A business that scales with the value of intelligence — OpenAI’s framing on scaling AI access and building sustainable business models.
Our approach to advertising and expanding access to ChatGPT — OpenAI outlines principles and plan to test ads for Free/Go tiers in the U.S.
Introducing ChatGPT Go, now available worldwide — Lower-cost ChatGPT tier aimed at wider global access.

See your AI credits usage history — Figma: admins/users get AI credits usage history (Jan 12 entry).
Enterprise key management now covers AI credits — Figma: Enterprise Key Management extends to AI credits (Jan 15 entry).

OpenAI partners with Cerebras to help developers build more reliable and scalable AI — Partnership positioning around compute, performance, and scaling.
GPT-5.2-Codex is now generally available in GitHub Copilot — Copilot adds GPT‑5.2‑Codex model selection broadly (plan-dependent).
GitHub Copilot CLI: Enhanced agents, context management, and new ways to install — Copilot CLI ships built-in agents (Explore/Task/Plan/Code-review) and better context controls.
Agentic memory for GitHub Copilot is in public preview — Copilot preview feature to retain/recall task context over longer agentic sessions.
GitHub Copilot now supports OpenCode — Adds support for OpenCode workflows (plus a related security advisory below).

Deploy AI agents on Amazon Bedrock AgentCore using GitHub Actions — Reference CI/CD pipeline for agent deployment with OIDC and security scanning.

Microsoft January 2026 Patch Tuesday fixes 3 zero-days, 114 flaws — Patch Tuesday overview with counts and zero-day notes.
NVD: CVE-2026-22813 — OpenCode: CVE entry to review and map to your internal exposure if using affected versions.

Introducing OpenAI for Healthcare — HIPAA-focused products for healthcare orgs (enterprise rollout + compliance framing).
Introducing ChatGPT Health — Dedicated ChatGPT experience for health/wellness, designed to incorporate personal health info securely.

Canva Shield: Safe, fair and secure AI — Updated policy/controls positioning for safe AI use (privacy controls + enterprise indemnification notes).
Updated left navigation bar — Figma UI update to left nav for quicker access to key areas (Jan 7 entry).
“Halftone”, “nostalgia” and more: 2025’s top Firefly prompt trends — Adobe prompt-trend snapshot you can use for internal creative prompting.

Gemini 3 Flash is now available in GitHub Copilot — Adds Gemini 3 Flash as a selectable Copilot model (availability varies by plan/policy).

OpenAI API Changelog (Jan 9): image edit fidelity fix — Fix for image edit fidelity behavior in gpt-image-1.5 / chatgpt-image-latest via /v1/images/edits.

Patch Tuesday, January 2026 Edition — Independent security commentary and highlights on Microsoft’s January patch set.

New collaboration controls — Figma: expanded collaboration controls (Jan 3 entry; helpful for governance in design workflows).
New workshop topics and templates in FigJam — Figma: new FigJam workshop content (Jan 2 entry; useful for AI-enabled ideation sessions).