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): Claude's Cowork feature — where you hand Claude a multi-step task and it works across your files, calendar, and other connected tools — has expanded beyond desktop to mobile and web, so a task started at your desk can be checked on from your phone. Anthropic also opened a government-focused version of Claude in beta this week.
- OpenAI (ChatGPT): Free-tier users are being moved to a new default model with unlimited text chats and an optional "Think" button for harder questions; paid users get a similar effort slider. Separately, OpenAI is retiring its Atlas browser tool on August 9 — if you used it, export any saved logins or bookmarks before they're gone. ChatGPT also logged a brief bout of elevated errors this week affecting some conversations, which OpenAI has since resolved.
- Microsoft (Copilot): Microsoft 365 Copilot is rolling out the ability to choose between OpenAI's and Anthropic's models directly inside Copilot Chat, plus new agents that can act inside Word, Excel, and PowerPoint without leaving the chat window. Worth knowing: Microsoft has also enabled OpenAI as a data subprocessor for Microsoft 365 by default — if that matters for your firm's client-confidentiality policies, check your tenant's toggle.
- xAI (Grok): xAI opened a public beta of "Grok Bot," an autonomous task-completing agent, with a further model update expected later this week.
Nothing notable to report on Google Gemini this week for a small-business audience.
Beyond the Platforms
EU's AI transparency rules take effect
As of August 2, businesses using AI systems that interact with the public — chatbots, AI-generated marketing content, deepfake-adjacent media — must disclose that fact under the EU AI Act's Article 50, with fines up to €15 million or 3% of global turnover for violations. Separately negotiated changes pushed the tougher "high-risk" system rules (covering things like automated credit scoring) back to December 2027.
Why it matters: if you serve any EU-based clients or run AI-driven chat or marketing tools that EU customers interact with, disclosure obligations may already apply to you.
Readers can't tell AI fiction from human writing — and prefer it
A study published this week in the journal Judgment and Decision Making found readers rated AI-generated short stories higher than human-written ones for quality and engagement, and only 40% of readers could correctly identify which was which.
Why it matters: if your business publishes any AI-assisted content, don't assume readers will notice or penalize it on quality grounds — the more relevant question is whether it's accurate and on-brand.
Industry Spotlight
For Legal Practices
Firms are adopting AI faster than they're measuring it
New BARBRI research finds law firms are rolling out AI tools more quickly than they're tracking the resulting changes in how lawyers actually work — a gap that leaves firms unable to show clients, insurers, or regulators what's actually improved.
Why it matters: if you can't measure the impact of the AI tools you've adopted, you can't defend the investment — or catch a workflow that's quietly introducing risk.
A new legal research tool bets against the "single answer" trend
While most legal AI research products now aim to hand lawyers one synthesized answer, litigation tech company Align launched a tool this week that instead returns the underlying cases themselves, for lawyers who want to do their own reasoning rather than trust a summary.
Why it matters: as AI research tools proliferate, it's worth matching the tool to how much independent verification your practice wants built into the workflow.
Legal-tech consolidation is accelerating
At least five acquisitions among legal AI vendors were announced in the past two weeks, including larger platforms absorbing smaller AI research and litigation-intelligence startups.
Why it matters: if you've adopted a niche legal AI tool, consolidation raises the odds it gets folded into a larger platform, changes pricing, or gets discontinued — worth checking your vendor's stability before renewing.
For Financial & CFO Firms
AICPA reminds members that existing ethics rules already cover AI
The AICPA published guidance this week clarifying that its existing Code of Professional Conduct — competence, due care, confidentiality, and independence — already governs how CPAs use AI tools; no separate AI-specific rulebook is coming.
Why it matters: if your firm has been waiting for formal AI rules before writing a use policy, the existing conduct code already applies — there's no reason to wait.
IRS disclosure rules for AI-assisted tax prep remain unclear
Tax professionals and the AICPA are pressing the IRS for clearer guidance on whether preparers must tell clients when AI tools were used to prepare their returns — current rules don't directly address it.
Why it matters: if you're a client of a tax preparer, or a CFO overseeing outside tax work, this is a good week to simply ask whether AI is part of that workflow.
New scorecard flags finance as the top target for AI-assisted identity fraud
A new industry analysis found financial firms face more AI-assisted identity fraud activity than any other sector, but are also better prepared to detect and document it than most industries.
Why it matters: fraud attempts against finance functions are increasingly AI-generated — a good prompt to double-check your firm's identity-verification controls hold up against synthetic documents and deepfaked voices, not just old-style forgeries.
On the Horizon
Analysts are warning that a chunk of AI agent projects will be abandoned within the next year. Gartner projects over 40% of "agentic AI" initiatives — tools that act autonomously on multi-step tasks — will be canceled by the end of 2027 due to unclear ROI and weak governance, and Forrester expects enterprises broadly to defer roughly a quarter of planned AI spending into 2027 for the same reasons.
Why get ahead of this now: before investing heavily in an autonomous AI tool, define upfront what "working" looks like and how you'll measure it — firms that skip this step are the ones most likely to abandon the investment later.
The EU's toughest AI rules — covering systems like automated credit decisions — won't bite until December 2027, later than originally planned, per this month's regulatory confirmation.
Why get ahead of this now: the extra runway is a gift, not a reason to wait — building compliant processes now, while the deadline is still distant, is far cheaper than retrofitting them under pressure later.
Try This This Week
Next time you have AI draft something with numbers or citations in it — a client-ready summary, a brief, a set of projections — add one line to your prompt: "List every figure, date, and citation you just used, separately, so I can check each one." It takes AI a few seconds and takes you a minute or two to scan, and it turns "trust the output" into "verify the output," which is the habit every current legal and accounting guidance on AI keeps pointing back to.