02 WEEKLY BRIEFING • SEPTEMBER 21, 2026

Technology
& Law Weekly.

The developments worth knowing before you practice.

CURATED BY THE
TECHNOLOGY LAW STUDENT ASSOCIATION

EDITOR'S NOTE

Legal AI is becoming a governed practice.

This week brought a new legal-AI platform, but the more interesting signal is the control layer around it. Courts are still punishing unverified use. Platforms are being pressed to address synthetic harm. Regulators are asking who independently tests the systems. The work is getting more capable, and the accountability is getting more concrete.

ISSUE 02

Five articles from this week.

LEGAL RESEARCH

OpenAI launches Astra for Law, with legal search and firm-specific controls.

OpenAI introduced Astra for Law, a configuration of GPT-6 Astra for law firms and legal-technology companies. The offering combines legal analysis instructions with a search index spanning U.S. cases, statutes, regulations, court rules, and administrative decisions. It will initially be available to selected firms through Trusted Access, with API access to follow.

Why it matters: The legal-AI race is moving past a generic chatbot in a lawyer costume. Search grounding, permissions, ethical walls, and reviewable workflows are becoming part of the product itself.

Read OpenAI's announcement ↗

ADOPTION

OpenAI is reportedly pairing its models with legal engineers who work inside firms.

Business Insider reports that OpenAI named Telon, a London legal-AI services startup, a select partner. Telon places legal engineers, many of them former lawyers, inside law firms and corporate legal teams to configure models, build agents, and train users around actual workflows.

Why it matters: Buying a tool and changing how work gets done are very different things. The emerging legal-AI job market will reward lawyers who can translate practice needs into durable, supervised systems.

Read Business Insider ↗

PROFESSIONAL RESPONSIBILITY

The New Mexico Supreme Court held a lawyer in contempt over an AI-generated appellate brief.

In a September order, the court found that attorney Stephen D. Aarons submitted a brief containing fabricated witness testimony and misrepresented legal authority after using ChatGPT. The court held him in direct contempt, struck the briefing, imposed a $5,000 sanction, referred the matter to the disciplinary board, and barred him from appearing before the court pending that process.

Why it matters: Verification is not a disclaimer you add after the fact. It is a professional task that must happen before a lawyer signs and files work produced with AI assistance.

Read the court's order ↗

SYNTHETIC MEDIA

Meta's Oversight Board calls for stronger action on an AI-generated video targeting a campaigner.

In a September 17 decision, the Oversight Board said Meta should remove an AI-generated video of a female health campaigner and criticized the company for leaving the content up. The Board also recommended broader policy changes, including clearer rules for manipulated imagery and measures to reduce the spread of high-risk AI-generated content.

Why it matters: Deepfake governance is becoming less theoretical. The legal questions now sit at the intersection of platform duties, privacy, political speech, evidentiary authenticity, and remedies that arrive after the damage has already traveled.

Read the Oversight Board decision ↗

AI GOVERNANCE

California asks how independent AI oversight could work in practice.

Governor Gavin Newsom's September 18 executive order accelerates implementation of California's new independent-verification and AI-auditor frameworks. It also directs state officials to assess whether frontier developers should face onsite independent evaluations, verified safety reporting, broader incident reporting, and a tested emergency shutoff mechanism.

Why it matters: Lawyers advising AI companies may soon spend less time arguing about high-level principles and more time designing evidence that a safety program, audit, or incident response actually works.

Read California's announcement ↗

WHAT TO WATCH

What counts as a real control when AI enters legal work?

Watch for a sharper distinction between a feature and a safeguard. A retrieval system, access setting, disclosure, audit, or human-review step only matters if someone can show when it applies, who owns it, and what happens when it fails.

TLSA QUESTION OF THE WEEK