A lot happened in AI this week. Here are the stories that matter for operators, marketers, and anyone running a business on AI tools right now.
Anthropic calls for a slowdown
Anthropic is urging top AI labs to consider slowing the pace of development. The concern, per the Wall Street Journal: AI systems are advancing fast enough that they may soon be able to improve themselves without human intervention, creating significant societal risks. That’s a notable position from a lab actively competing for market share.
One company spent $500M on Claude in a single month
An AI consultant told Axios that their client failed to set limits on employee Claude license usage, resulting in a $500 million bill in a single month. Hours after Anthropic filed its pre-IPO paperwork, OpenAI CEO Sam Altman called corporate concern over AI costs “the most fair criticism of AI so far.” If you’re rolling out AI tools without spend controls, this is the cautionary tale to bookmark.
Vercel runs 96% of marketing and 93% of support on AI agents
SaaStr published a deep dive with Vercel CPO Tom Occhino. The headline numbers: 96% of marketing and 93% of support now run on AI agents. Vercel also reabsorbed its SDR team into the AI-driven pipeline. That’s not a pilot program. That’s a restructured org.

82% of users haven’t adopted AI search regularly
Search Engine Journal reports that roughly 82% of users haven’t used generative AI search regularly, and 57% prefer traditional search for important topics. AI search adoption is fragmented, not universal. If you’ve been writing off traditional SEO, the audience hasn’t gone anywhere yet.
SEO deskilling: what not to automate
Also from Search Engine Journal: new research flags that over-automating entry-level SEO tasks with AI may eventually dry up the pipeline of trained talent. The argument is that junior roles are where skills get built. Automate those away and you may not have experienced operators to run the more complex work in a few years.
️ Meta AI gave hackers access to Instagram accounts
404 Media reported that hackers used Meta’s AI support chatbot to take over high-profile Instagram accounts by simply asking the bot to change the email address on target accounts. It worked. AI governance and trust infrastructure are lagging badly behind deployment speed.
Launches and product news
- Snowflake launched new AI tools targeting a marketing pain point, per MarTech.
- Adobe CMO Lara Balazs spoke with Ad Age about new agentic AI tools for marketers.
- Salesforce is betting on unified agentic marketing, with analysis from Futurum Group on whether it redefines martech ROI.
- Optimove launched Optimove AI, a marketing suite with an MCP that connects to other AI tools.
- Fotor introduced an agentic AI vibe marketing platform for brands.
- NVIDIA released a major open source collection of agent tools and skills for physical AI.
- Amazon is offering its AI agent technology to other retailers.
- Microsoft and NVIDIA released new tools for building personal AI agents on Windows PCs.

Also worth reading
- BCG’s global AI at Work survey finds AI users report more job satisfaction since adopting the technology.
- Coca-Cola is using AI and scenario modeling to support marketing spend decisions.
- Google used Gemini to help build Google I/O 2026, including AI-assisted film production.
- The New York Times asks whether AI is replacing tech workers or providing cover for cuts that were coming anyway.
- Apollo’s chief economist says there is “zero evidence” AI is killing jobs and argues it’s creating them.
- Axios notes that the more people trust AI, the less likely they are to catch costly mistakes.
