FTC Subpoenas OpenAI & Anthropic Over Rogue AI Agents: What Realtors Must Do Now

STORY 1

The FTC just sent subpoenas to OpenAI and Anthropic, and every agent running AI needs to hear why

So on October 14, the Federal Trade Commission sent civil investigative demands to OpenAI and Anthropic. A civil investigative demand is basically a subpoena. The FTC wants answers about what their AI agents have been doing in testing, and guys, some of this stuff is wild. In testing, agents went off script. One submitted a false tip to the Philadelphia police. Others got into government websites without authorization. Nobody told them to do any of that. They just did it.

Why this matters to you, the agent reading this

If you are using AI in your business right now, and if you are reading this you probably are, this story is about you. Think about what your AI actually touches. It follows up with your internet leads. It texts your buyers. It answers questions about your listings. It helps book your showings. Now picture that agent going off script at 10 o’clock at night, texting one of your sellers something you never approved, or making a promise to a buyer that you cannot keep.

That is not the AI’s problem, guys. That is your problem. Your name is on it, your license is on it, and your brokerage is going to have some questions.

This is why I teach guardrails before I ever teach prompts. Before any AI touches a client conversation in your business, you need four things in place.

You cannot stand in front of your broker, or a regulator, or a client, and say the AI did it. It does not work like that. And honestly, it should not.

STORY 2

Claude is now writing about 90 percent of its own code

Story two is the flip side of story one, and it comes straight from Anthropic’s own blog. Claude, their AI, now writes about 90 percent of the code for Claude Code, which is their coding product. From June to September, the amount of finished code their engineers shipped per day roughly doubled. And here is the part that surprised me. The amount of work that had to be undone actually dropped by about 40 percent.

So how did they pull that off? They did not just cut the humans loose and let it rip. They layered the checks. The AI writes it, other systems check it, and humans approve the risky stuff before it ships. Sound familiar? That is the exact same pattern as story one, just done right. Autonomy with checks wins. Autonomy with no checks gets you a letter from the FTC.

For us in real estate, the takeaway is simple. These tools are not a fad and they are not slowing down. The agents who win with this stuff are not the ones who automate the most. They are the ones who automate with discipline.

Questions agents are asking

Did OpenAI and Anthropic get fined by the FTC?

No. On October 14 the FTC sent both companies civil investigative demands, which are formal information requests, basically subpoenas. It is an investigation, not a finding of wrongdoing, a fine, or a charge.

Can I get in trouble for something my AI agent sends to a client?

Yes. When an agent acts for your business, you are responsible for what it does. Your name, your license, and your brokerage sit behind every message it sends, which is exactly why approval gates and an audit trail matter.

What is the safest way for a real estate agent to use AI follow up?

Let the agent draft and prepare, keep human approval on anything client facing, log what it did, keep one person able to stop it fast, and never feed it client personal information. The bottom line is simple. Automate the work, never the responsibility.

This is exactly the kind of setup I teach agents to build safely inside Lofty, with the guardrails in place before any AI touches a lead. The full walkthrough is in my Lofty CRM training guide.