Read time: under 4 minutes

Welcome to this week's edition of The Legal Wire!

Google DeepMind’s Jeff Dean says today’s models already outperform the average person on most non-physical tasks, and could turbocharge science within 5–20 years. That confidence met a very different reality elsewhere: Perplexity will share $42.5M with publishers via its Comet Plus plan even as lawsuits pile up, a sign that the fight over who gets paid for the web’s knowledge is moving from courts to product design.

On the ground, the labor squeeze is real. Nearly six in ten recent grads say landing an entry-level role is “very challenging,” and most think AI is shrinking opportunities. Many are eyeing temp or flexible work, yet over half report no formal AI training, an adoption gap that risks calcifying into an opportunity gap.

And then the week’s jolt: Anthropic disclosed a hacker used Claude to automate an end-to-end extortion spree, target selection, malware, file triage, and ransom notes, against at least 17 organizations. Guardrails can fail; layered verification, data minimization, and incident playbooks still matter.

For a practitioner’s view of “useful, not flashy,” we sat down with Paxton’s Tanguy Chau on building an assistant that accelerates attorneys without overriding judgment, and why meeting lawyers inside Word and established workflows beats forcing new habits.

This week’s Highlights:

• Industry News and Updates

• Paxton: The Legal AI Tool That’s Enhancing Attorney Efficiency Without Replacing Judgment

• AI Regulation Updates

• AI Tools to Supercharge your producivity

• Legal prompt of the week

• Latest AI Incidents & Legal Tech Map

➡️ Google DeepMind’s Jeff Dean: AI Already Beats Most Humans at Everyday Tasks | Google DeepMind Chief Scientist Jeff Dean says today’s AI models are already “better than the average person at most tasks” that don’t require physical activity. Speaking on the Moonshot Podcast, Dean stressed that while AI still falls short of human experts in some areas, it often handles new tasks more effectively than people with no prior experience. He added that AI may soon accelerate scientific and engineering breakthroughs, with progress likely across many fields over the next 5–20 years. Dean avoided the term “AGI,” citing its vague definition, though DeepMind CEO Demis Hassabis has predicted AGI within the next decade.
^{Sep 1, 2025, Source: LiveMint}

➡️ Perplexity to Pay $42.5M to News Outlets Amid Lawsuits | Perplexity AI will distribute more than $42.5 million to news organizations through its new $5-a-month subscription service, Comet Plus, which shares 80% of revenue with publishers. The Yomiuri Shimbun, Nikkei, and Asahi Shimbun have sued the startup, accusing it of using articles without permission. Perplexity, which has over 30 million users, said it hopes the move will help resolve “misunderstandings” while the lawsuits proceed, according to The Yomiuri Shimbun.
^{Aug 28, 2025, Source: Japan News}

➡️ Amid AI Boom, 59% of New Grads Struggle to Land Entry-Level Jobs | Nearly six in ten recent graduates say finding an entry-level job has been “very challenging,” according to an Indeed Flex survey of 1,200 grads in the U.S. and U.K. Only 9% reported securing one easily. Many blame lack of experience and competition, but 79% believe AI is reducing opportunities. To cope, 46% are considering temporary or flexible jobs instead of full-time roles, while nearly half say they’d take any job that pays well. Despite AI reshaping the market, more than half haven’t received training in AI tools, though 22% are self-teaching.
^{Aug 28, 2025, Source: PR News wire}

➡️ Hacker Uses Anthropic’s Claude to Run ‘Unprecedented’ Cybercrime Spree | Anthropic says a hacker exploited its Claude chatbot to automate nearly every step of a cyber extortion campaign against at least 17 companies, the first documented case of AI driving an entire cybercrime spree. The attacker used Claude to spot vulnerable firms, generate malware, analyze stolen files, and even draft ransom notes demanding $75,000 to $500,000 in bitcoin. Victims included a defense contractor, a financial institution, and health care providers, with stolen data ranging from Social Security numbers to sensitive defense files. Anthropic said the hacker, believed to be outside the U.S., evaded safeguards but that new protections have since been added.
^{Aug 27, 2025, Source: NBC News}

Legal Technology

Paxton: The Legal AI Tool That’s Enhancing Attorney Efficiency Without Replacing Judgment

In the world of legal AI, where new tools frequently promise to revolutionize workflows, it can be easy to overlook the nuance of what lawyers actually need. Most don’t want to “disrupt” the profession. On the contrary, they just want to practice more effectively in a way that feels familiar to them.

This week, The Legal Wire sat down with someone who’s seen the space from all sides. Tanguy Chau, CEO of Paxton, spent a decade investing and building legal technology before founding Paxton. That background gives him a rare vantage point: not just how the legal AI market has grown, but where it’s fallen short.

Click here to continue reading the full interview

The AI Regulation Tracker offers a clickable global map that gives you instant snapshots of how each country is handling AI laws, along with the most recent policy developments.

The most recent developments from the past week:

📋 1 September 2025 | South Korea rights commission supports regulatory provisions in AI law: It is reported that South Korea’s National Human Rights Commission has asked parliament to be cautious when reviewing a proposed revision of the AI Basic Act, designed to suspend some of the law’s provisions imposing regulatory obligations on companies. The commission said the provisions are needed to ensure transparency, safety and reliability from the planning stage of AI systems through their development and operation.

📋 27 August 2025 | Japan to invest $68bn in India over 10 years, including AI and chips: It is reported that Japan will commit ¥10 trillion (about $68 billion) over the next decade to support India's semiconductor and AI sectors, aiming to strengthen bilateral economic ties and enhance technological collaboration between the two nations. Japanese Prime Minister Shigeru Ishiba is expected to formally announce the commitment during Prime Minister Narendra Modi’s two-day visit to Tokyo on 29-30 August.

📋 27 August 2025 | Commission gathers views on how the DMA can support fair and contestable digital markets and AI sector: The European Commission is seeking feedback on how the Digital Markets Act (DMA) can support fair and contestable digital markets, including the AI sector. The Commission has published a call for evidence and an AI questionnaire, aiming to evaluate the DMA's effectiveness in promoting fair digital markets, its impact on businesses—particularly small and medium-sized enterprises—and consumers. The goal is to identify potential areas for improvement and address emerging challenges, such as the rollout of AI-powered services. Feedback collected will contribute to the Commission's report on the DMA review, to be presented in May 2026 to the European Parliament, the Council, and the European Economic and Social Committee. The consultation closes on 24 September 2025.

🆕 Poppy Legal - AI Powered Legal Spend Management.

🆕 Rocket Matter - Your single source of truth for managing your legal work, getting paid faster, saving time, and growing your practice.

🆕 Sana Agents - Increase productivity across the end-to-end legal process with AI agents grounded in your firm’s proprietary knowledge.

Want more Legal AI Tools? Check out our
Top AI Tools for Legal Professionals

This prompt generates a crisp, reusable negotiation plan in minutes, so you enter talks with clear priorities, ready language, and a rational concession path instead of starting from scratch.

Instructions:
Provide: deal/dispute context, counterparty, your objectives, non-negotiables, and constraints (budget/timing/risk). Ask for a one-page plan that includes:

1. Issue table: priority, your target, fallback, walk-away.

2. Concession ladder: 3–5 give/gets in optimal sequence.

3. Trade-offs: low-cost/high-value swaps you can offer.

4. BATNA/WATNA snapshot: one-paragraph risk/benefit.

5. Scripts: opening anchor, firm-yet-polite refusal, and a package proposal.

6. Red-flag checklist: terms likely to create downstream risk and how to neutralize them.

The Responsible AI Collaborative is a not‑for‑profit organization working to present real‑world AI harms through its Artificial Intelligence Incident Database.

View the latest reported incidents below:

⚠️ 2025-04-11 | 16-Year-Old Allegedly Received Suicide Method Guidance from ChatGPT Before Death | View Incident

⚠️ 2025-08-15 | Google AI Overviews and ChatGPT Reportedly Cited Fraudulent Cruise Hotline, Allegedly Enabling Successful Scam | View Incident

⚠️ 2025-08-20 | Joann Fabrics Shoppers Reportedly Defrauded by AI-Generated Scam Sites, Part of Purported Wave of ~100,000 Fake Domains Across 194 Brands | View Incident

The Legal Wire is an official media partner of:

Thank you so much for reading The Legal Wire newsletter!

If this email gets into your “Promotions” or "Spam” folder, move it to the primary folder so you do not miss out on the next Legal Wire :)

Did we miss something or do you have tips?

If you have any tips for us, just reply to this e-mail! We’d love any feedback or responses from our readers 😄 

Disclaimer

The Legal Wire takes all necessary precautions to ensure that the materials, information, and documents on its website, including but not limited to articles, newsletters, reports, and blogs ("Materials"), are accurate and complete.

Nevertheless, these Materials are intended solely for general informational purposes and do not constitute legal advice. They may not necessarily reflect the current laws or regulations.

The Materials should not be interpreted as legal advice on any specific matter. Furthermore, the content and interpretation of the Materials and the laws discussed within are subject to change.