AI News Roundup: A Shared AI-Training Opt-Out, Banks Block Agents, and a Funding Squeeze

Four AI labs agree on one robots.txt signal for training opt-outs, big banks stop outside AI agents from logging in, OpenAI trims its funding round, and New York limits chatbots that talk to kids.

Here is the AI news from the last day or so that matters for small businesses using AI agents. Every item links to its source.

One opt-out for AI training. OpenAI, Google, Microsoft and Anthropic say they will honor a shared robots.txt token that tells their crawlers not to train on a site's pages. Anthropic supports it now, OpenAI says "in the coming months," and Microsoft names February 1, 2027. If you publish content, it's worth a look at your robots.txt. (AI Pro Playbook, DailyAI Briefing)

Banks block outside AI agents. JPMorgan Chase stopped letting third-party AI agents log in to customer accounts with a customer's password, and Citigroup and Bank of America followed. Agents that act for you still need sanctioned access, not shared logins. (AI Lately)

The money tightens. OpenAI reportedly cut its funding round target from $30 billion to about $20 billion, and Blackstone now expects Anthropic's debt to be rated junk after its IPO. Intel fell about 29% on a weak forecast tied to AI-driven memory shortages. (AI Edge Briefing)

New York limits chatbots with kids. A new law bars AI chatbots from pretending to be human while talking to a child, with fines of $25,000 per violation starting January 1. (AI Lately)

What it means for you: rules around AI are getting clearer, and honesty wins. A voice agent that answers fast, says what it is, and hands hot leads to a person is the safe bet. That is how Ozkra works. Want to see it on your calls? Talk to us.