This briefing is informational only and does not constitute legal or financial advice. It reports on public industry developments and does not reference any Tienta client.
This Week in AI
Platform Roundup
- Anthropic (Claude): Code Review for Claude Code left research preview on October 15 and is now available to Team and Enterprise organizations. Anthropic says most reviews finish in 6 to 10 minutes, down from about 20, and admins choose Quick, Standard (the default), or Deep review for each repository. Eligible Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027, and the credits expire 30 days after they are claimed.
- OpenAI (ChatGPT): Slides began rolling out in ChatGPT on October 16 for eligible Pro, Business, and Enterprise users who have access to Spaces. You ask ChatGPT in Work mode to build a presentation inside a Space. Coverage reports co-editing and export to PowerPoint, PDF, and Google Slides, but those details come from secondary reporting.
- Google (Gemini): Gemini 4 Argon, which Google first released on September 30 to a limited group of cyber defenders, became generally available on October 13. It is also generally available in GitHub Copilot on the Pro+, Max, Business, and Enterprise plans.
Beyond the Platforms
The AICPA is piloting a certification for AI agents, delivered by CPA firms
On October 19, the Journal of Accountancy reported that the AICPA is partnering with the Artificial Intelligence Underwriting Company on a single engagement that combines a SOC 2 report with AIUC-1 certification, which it describes as the first standard for AI agent safety, security, and reliability. The SOC 2 portion is an attestation examination performed by a licensed CPA under AICPA standards. The AIUC-1 portion is a certification. Several CPA firms are already delivering initial engagements, and a full launch is expected in early 2027. Both require annual renewal, and AIUC-1 adds quarterly technical retesting.
Why it matters: sooner or later your clients, and your own customers, will ask how you know your AI is safe to rely on. This is an early sign of what an answer might look like: an independent report rather than a vendor's assurance. It is a pilot, not an established standard, so treat it as a direction to watch.
Industry Spotlight
For Legal Practices
The federal evidence rules committee took up deepfakes and AI-generated evidence on October 15
The Advisory Committee on Evidence Rules, which drafts changes to the Federal Rules of Evidence, met in Boston on October 15 with panels of experts on two topics. The first is a draft new Rule 901(c) on deepfakes. As described in the committee's agenda book, the party challenging an item would have to present evidence sufficient to support a finding that it was fabricated. If they did, the party offering it would have to show it is more likely than not authentic, a higher bar than the usual standard for authenticating evidence. The agenda says the committee has worked on this for seven meetings and has prepared the rule but not finally approved it. The second topic is proposed Rule 707, which covers AI output offered as evidence without an expert. The committee withdrew the version it had published for comment at its spring meeting and is studying a modified one. This briefing does not report an outcome, because the agenda was published ahead of the meeting.
Why it matters: neither rule is in force, and neither is close to it. But the questions are live now: any firm that handles disputes will meet claims that a recording or image is fake, and a court that applies a stricter standard will want to see how the item was obtained and kept. Preserving originals and their metadata, and knowing who handled them, costs little today and is hard to reconstruct later.
For Financial & CFO Firms
The AICPA issued an AI practice aid for financial statement auditors
Published October 13, "Artificial Intelligence (AI) Considerations for Financial Statement Auditors" has two parts. The first covers how an organization's use of AI in financial reporting or controls affects the audit, including risk assessment, data quality, the audit team's capabilities, and AI-specific risks such as bias and hallucinations. The second covers how audit firms use AI tools on engagements, including quality management, supervision, and partner responsibilities. The AICPA says more guidance on risk of material misstatement and audit evidence will be added to the first part later. Its chief auditor, Jennifer Burns, said, "Ultimately, the auditor remains responsible for the audit and the audit opinion."
Why it matters: if you prepare financial statements, expect your auditor to ask where AI touches the numbers, how the data feeding it is controlled, and who reviews the output. If you are the auditor, the responsibility stays with you whether the work was done by a person or a tool. Either way, a short written inventory of where AI is used in financial reporting is worth having before the questions arrive.
For SaaS & Engineering Teams
A new frontier model arrived in GitHub Copilot already turned on
GitHub's changelog for October 13 says Gemini 4 Argon is generally available in Copilot and rolling out gradually. Business and Enterprise administrators can manage access through the model policy in Copilot settings, and under default model enablement, new models are switched on automatically unless an administrator has turned off the global default or explicitly disabled that model. GitHub says the model is billed at the provider's list pricing under usage-based billing. The same week, Anthropic's Code Review became generally available, but an admin has to enable it before it reviews anything.
Why it matters: the two products take opposite approaches. One needs an admin to opt in, and the other opts everyone in unless an admin opts out. Neither is wrong, but a team that does not know which of its tools work which way will find out when a new model starts handling private code and its usage shows up on the bill. Someone should own the setting for each tool.
On the Horizon
The 10th Circuit's comment period on its proposed AI filing rule closed October 18. As covered in issues 7 and 10, the proposal would require a certification that a person reviewed any filing drafted with generative AI. No adoption has been reported yet. If adopted, it would join California's SB 574 and a growing list of state court rules on AI verification.
Why get ahead of this now: the common thread across these rules is verification, not disclosure. A firm whose process already requires a named person to check every citation before filing is ready for whichever version is adopted.
Try This This Week
Pick the three AI-enabled tools your team uses most, such as your code editor assistant, your chat tool, and your document platform. For each, find out who can switch on new models and features, and whether new ones turn on automatically. Write the answer next to the tool's name. If nobody knows, that is your first task.