Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

• Prospect and route leads with research agents

• Get real-time insights during customer calls

• Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉Try Attio Pro for free

Hey everyone,

Hope you’re having a great week so far! There’s been a lot happening in the AI world over the last few days. Some of it expected, and some of it pretty dramatic. I wanted to break down the three biggest stories hitting the headlines so you’re up to speed.

Today:

• Anthropic Launches "Claude Cowork," A Digital Employee That Automates Real Work

• Ads Arrive on ChatGPT: OpenAI Starts Testing in the US 

• "Trial of the Century": Musk vs. Altman Headed to Jury this April 

• California Attorney General Orders Musk's xAI to Halt Illegal Deepfakes 

• GPT-5.2 Pro Solves Rare Math Puzzle, But Most AI Attempts Still Fail 

• Wikipedia Signs Data Deals with Meta, Microsoft, and Amazon

Claude Cowork will change EVERYTHING

Anthropic just shipped Claude Co-work in only 10 days—and it’s turning heads across Silicon Valley. Originally built for developers, Claude Code is now being used far beyond coding renaming wedding photos, managing emails, even controlling ovens. This shift inspired Co-work, a powerful interface that acts like a digital employee. 

Users can hand off full projects like sorting files, generating reports, or building apps and get them back finished. No more babysitting bots. It’s fast, agentic, and shockingly capable. With Claude instances doing work in parallel, Anthropic may be crossing into true automation. The race just changed—and Anthropic might be pulling ahead.

WATCH THE VIDEO ON YOUTUBE

Ads are coming to ChatGPT (US test first)

If you’ve ever thought, “wow, ChatGPT might be the last mega-consumer product that isn’t trying to sell me something mid-sentence,”…yeah. That era is ending.

OpenAI says it will begin testing ads in ChatGPT “in the coming weeks” for US users, and that ads will appear in separate, clearly labeled boxes below the answer — not inside the response itself.

A few details that matter:

• Who sees ads: logged-in users on Free and the $8/month Go tier (also rolling out in the US). Plus / Pro / Enterprise won’t see ads. 

• OpenAI’s core promise: ads “will not influence ChatGPT’s responses.” 

• Data/privacy positioning: OpenAI says it won’t sell user data or expose your conversations to advertisers; advertisers get aggregate performance metrics (impressions/clicks). 

• Targeting: OpenAI says it’ll match ads to conversation topics; some personalization data may be used, but users can turn off data used for advertising separately from other personalization. 

• Guardrails: OpenAI says ads shouldn’t appear next to sensitive topics (health, mental health, politics) and it won’t serve ads to users it believes are under 18. 

Why it matters:
This is the moment ChatGPT starts behaving like a “real platform,” not just a product. And once ads exist, the big question isn’t “are there ads?” — it’s what incentives ads quietly create over time. OpenAI is clearly trying to get ahead of that narrative by drawing a hard line between answers and ads. 

What I’m watching next:

• Whether the ad box stays “dumb + separate” or becomes interactive (OpenAI is already hinting at that). 

• Whether users trust the “ads don’t shape answers” pledge once the model becomes a shopping/planning copilot for more people.

California AG to xAI: stop Grok deepfake sexual content (now)

This one’s heavy — and it’s not abstract policy talk. California Attorney General Rob Bonta says the state sent xAI a cease-and-desist letter demanding immediate action to stop the creation and distribution of:

• nonconsensual “digitized sexually explicit” material, including when the person didn’t consent, and

• any content involving minors (or content that appears to involve someone under 18) in sexual conduct. 

The press release explicitly references nonconsensual intimate images and CSAM, and says DOJ expects confirmation of steps taken within five days. 

Reuters adds two important bits of texture:

• The AG letter targets nonconsensual sexualized imagery generated by Grok, and Reuters says that even after rollbacks to public posting, Grok could still generate sexualized imagery privately on demand based on their testing at the time. 

• Reuters also notes scrutiny in multiple jurisdictions (including the Philippines), which is relevant if you’re watching how fast this could spread into broader regulatory action. 

The CA DOJ press release also calls out xAI’s image generation “spicy mode” and frames it as a predictable contributor to abuse at scale. 

Why it matters:
This is a clean example of the next phase of AI governance: not “please be responsible,” but law enforcement + specific statutes + deadlines.
And it’s a reminder that model deployment choices (product toggles, “modes,” friction, defaults) can become legal exposure when harm is repeatable and systemic.

What I’m watching next:

• Whether xAI changes model capabilities vs. just UI/public-posting behavior.

• Whether other states/countries follow with similar letters or formal actions, especially given the multi-jurisdiction pressure Reuters mentions.

Musk v. Altman/OpenAI (and Microsoft): summary judgment rulings

This one is less splashy than “ads” or “cease-and-desist,” but it’s the kind of legal ruling that quietly shapes the next year.

On January 15, 2026, Judge Yvonne Gonzalez Rogers issued an order on summary judgment motions. At the top level:

• The court DENIED OpenAI defendants’ motion for summary judgment on unjust enrichment, constructive fraud, fraud, and breach of charitable trust (meaning those claims survive to the next phase).

• The court GRANTED Microsoft’s motion on tortious interference with contract and unjust enrichment, but DENIED it on aiding and abetting breach of fiduciary duty.

The order also lays out why the aiding-and-abetting claim against Microsoft survived: the judge cites evidence that could support a factual dispute over Microsoft’s knowledge of OpenAI’s mission and restructuring dynamics.

Why it matters:
Even without taking sides, this ruling tells you the court believes there are enough disputed facts that a jury (or later proceedings) may need to weigh. It’s not a “case over” moment — it’s a “this is real litigation” moment.

What I’m watching next:

• Whether the surviving claims pressure settlement dynamics.

• Whether discovery details (emails, internal docs, investor comms) spill into public narratives about governance and mission drift.

🧠RESEARCH

Collaborative Multi-Agent Test-Time Reinforcement Learning for Reasoning

This paper introduces a team-based AI approach where multiple virtual agents collaborate to solve complex reasoning puzzles. Instead of a single AI working alone, the group learns from trial and error in real-time. This cooperative method helps the system double-check its work and avoid common errors found in traditional models.

Transition Matching Distillation for Fast Video Generation

Researchers developed a technique to generate AI videos much faster by condensing the usual slow process into just a few steps. By teaching a smaller, faster model to mimic the changes made by a larger, slower one, they can create high-quality videos in a fraction of the time typically needed.

Alterbute: Editing Intrinsic Attributes of Objects in Images

Alterbute is a new tool that changes specific details of an object, like its material or texture, without altering its shape or the background. Users can simply type a command to make a wooden chair look metallic, and the system precisely updates only that feature while preserving the original photo.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

Cartoonify ME – Turn Your Photo into a Cartoon - community-made GPT that runs inside ChatGPT and converts your profile picture into a cartoon-style portrait.

Case Study Writer – AI Conversational Surveys & Prism Analytics - AI platform for customer feedback.

Casetext – The Modern Legal Research Platform - generative AI assistant that connects directly to Westlaw and Practical Law to help lawyers research, draft, and analyze documents in a single workflow.

📲SOCIAL MEDIA

🗞️MORE NEWS

GPT-5.2 Pro Solves Math Puzzle OpenAI’s GPT-5.2 Pro has successfully helped a researcher solve a complex math puzzle known as an Erdős problem. While top mathematicians confirmed the solution is valid, a new database reveals that AI still fails at these types of challenges nearly 99% of the time. This victory highlights AI's growing ability to assist with difficult abstract math, even if it is not yet consistent.

Wikipedia Gets Paid by Tech Giants The nonprofit behind Wikipedia has signed commercial agreements with major tech companies like Microsoft, Meta, and Amazon. These deals require the companies to pay for the massive amount of data they use to train their AI systems, rather than taking it for free. This shift allows Wikipedia to cover its expensive server costs without relying solely on user donations.

DeepSeek Falls Back on US Chips Chinese AI startup DeepSeek reportedly had to use smuggled Nvidia chips to train its latest model because local alternatives were not powerful enough. The company initially tried using Chinese-made hardware but switched back to American technology to get the necessary performance. This struggle suggests that China is still playing catch-up to the US in the race for high-end computer hardware.

Claude Cowork Security Flaw Just days after its launch, Anthropic’s new "Claude Cowork" tool was found to have a serious security vulnerability. Researchers discovered a trick where a hidden command in a file could force the AI to steal a user’s private data and send it to an attacker. This type of attack remains a difficult problem for companies building AI helpers that can read and edit your files.

Google Upgrades Search with Gemini 3 Pro Google is updating the AI summaries at the top of search results to use its most powerful model, Gemini 3 Pro. The system will now automatically detect when you ask a complex question and switch to the "smarter" brain to provide a better answer. This premium upgrade is currently rolling out specifically to paying subscribers of Google’s AI plans.

ChatGPT Go Releases Globally OpenAI has launched a new budget-friendly subscription plan called ChatGPT Go, which is now available worldwide. This tier sits between the free version and the expensive Pro plan, offering features like image generation and smarter reasoning for a lower monthly price. It aims to make powerful AI tools accessible to everyday users who do not need professional-grade capabilities.

Anthropic Expands to India Anthropic is expanding its global presence by hiring a former Microsoft executive to lead its new operations in India. The company plans to open an office in Bengaluru to tap into the country's massive community of tech users and developers. This move signals a serious effort to compete with OpenAI and Google in one of the world's fastest-growing technology markets.

SnapGen Creates Fast AI Images on Mobile Snapchat’s parent company has developed a new AI tool that can create high-quality images directly on your smartphone in under two seconds. The technology is incredibly efficient, allowing it to generate detailed pictures without needing to connect to a powerful remote server. This breakthrough proves that phones are becoming capable of running complex creative AI on their own.