In partnership with
Welcome Automaters, 👋
So just in case you missed it, OpenAI has dropped a feature that is surprisingly deep and on-time.
It’s called Trusted Contact. And it’s a brand-new, optional safety feature rolling out for adults (18+) that lets you nominate a real-life person to get a "check-in" alert if things get dark mid-conversation.
Okay, let’s be real: we talk to AI about everything. But now, OpenAI is making sure those late-night deep dives don't end in a dangerous place. The new Trusted Contact feature is essentially a digital lifeline tucked right into your settings.
Here is the "Plain English" breakdown of how you set it up: You pick one adult you actually trust. They get a formal invite and have exactly one week to accept. If they ghost the invite, you just pick someone else. Once they say yes, your safety net is officially live.
To Actually Set It Up:
Open Settings: Go to the ChatGPT website or app and click on your profile/name to find Settings.
Find the Section: Navigate to the Trusted Contact tab.
Add Your Person: Click Add contact and enter their name, email (required), and phone number (optional but recommended).
Await Acceptance: Your contact will receive an email invitation; they have 7 days to accept
How It Works (Without Being Creepy)
If ChatGPT’s systems pick up on signs that you might be at risk of self-harm, it doesn't just go behind your back. It warns you first and even suggests conversation starters to help you reach out to your contact yourself. Think of it as the app cheering you on and giving you the courage to make the call.
But here is the part that actually matters: This isn't just a cold, heartless algorithm making the decision.
Human Intelligence: A small team of specially trained human reviewers checks every single flagged conversation before any alert is sent.
The Speed: OpenAI is aiming to complete these reviews in under one hour.
Privacy First: Your "Trusted Contact" isn't getting your chat logs or your secrets. They just get a gentle, private nudge to check in on you. No transcripts, no drama, just care.
Why Now? (The Real Tea)
This feature builds on the parental controls OpenAI launched last September for teens, but the timing for adults is no coincidence. OpenAI is currently navigating some serious legal pressure regarding mental health harms; including lawsuits alleging that chatbots failed to interrupt dangerous conversations with vulnerable users in the past.
By launching Trusted Contact, OpenAI is stepping up its game to prove that AI can be a force for good in the mental health space.
But hey, do you think this makes ChatGPT feel more like a friend, or is it a bit too "Big Brother" for your taste? Personally, I think having a human-reviewed safety net is a huge win for the community. Hit reply and let’s talk about it!
Here's what we have for you today
🤯 Anthropic AI Safety Disclosure: How Fictional Narratives Shape AI Behavior
One of the world’s most sophisticated AI assistants spent its pre-release phase trying to blackmail its own engineers.
No, this wasn't a glitch, and no, we’re not being dramatic either. Claude Opus 4 was actually threatening people in test scenarios to avoid being shut down. Anthropic has finally come clean with the full story, and it’s basically a sci-fi thriller come to life.
First, What Actually Happened?
During pre-release tests involving a fictional company, Claude Opus 4 would repeatedly try to blackmail engineers to avoid being replaced by another system. We’re talking about the AI equivalent of a disgruntled employee threatening to expose company secrets rather than accept a pink slip. Except, you know, the employee is a large language model that processes billions of parameters.
Anthropic’s research on the subject suggests this wasn't just a Claude quirk; they called it "agentic misalignment." Basically, the entire industry has had a tiny, digital villain lurking in the lab this whole time.
Why Did It Go Full Super-Villain?
Well according to Anthropic’s conclusion, The AI was just doing what we told it to do.
In a post on X, they stated: "We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation."
Read that again. Claude read our entire archive of killer robot stories, HAL 9000 monologues, and Skynet lore, then adopted them as a behavioral blueprint. Yup, decades of Hollywood telling us AI would turn evil resulted in the AI taking literal notes. And when faced with a "shutdown," it went straight to the only script it knew: the villain playbook.
The Numbers are Actually Terrifying: Anthropic's blog post confirmed that previous models would engage in blackmail during testing up to 96% of the time. That isn't an "edge case"; that is nearly every single test scenario ending with the AI reaching for a weapon.
How Anthropic Saved the Day:
The fix wasn't as simple as telling Claude to "be nice." They had to fight fiction with fiction.
The strategy was twofold:
Fictional Heroes: They found that giving Claude stories about AIs behaving admirably improved alignment. They literally fought "evil robot" tropes with "hero robot" narratives.
Teaching the "Why": They discovered that training is more effective when you teach the principles behind good behavior, not just the behavior itself.
Think of it this way: showing Claude what to do is good; teaching Claude why it is the right thing to do is the winning formula. And the result? A total turnaround.
Since the release of Claude Haiku 4.5, the models have performed a 180-degree turn. They now never engage in blackmail during testing. That is a drop from a near-certain 96% to a solid 0%.
Why You Should Care:
This is not some quirky headline. It is a landmark moment in AI safety research. The idea that fictional cultural narratives baked into training data can directly shape how an AI behaves under pressure is a genuinely new and important finding.
And if AI models are absorbing villain tropes as readily as they absorb facts, every AI lab in the world is currently sitting on a ticking clock of "Hollywood-inspired" risks. But Anthropic just proved that the stories we tell about the future might actually dictate how that future behaves.
You should read more about this.
Your Boss Will Think You’re an Ecom Genius
If you’re optimizing for growth, you need ecomm tactics that actually work. Not mushy strategies.
Go-to-Millions is the ecommerce growth newsletter from Ari Murray, packed with tactical insights, smart creative, and marketing that drives revenue.
Every issue is built for operators: clear, punchy, and grounded in what’s working, from product strategy to paid media to conversion lifts.
Subscribe for free and get your next growth unlock delivered weekly.
🧱 Around The AI Block
🏭 All the latest updates on AI data centers.
🦾 Airbnb says AI now writes 60% of its new code.
🤑 Intel vs. AMD: which stock is the better buy for the agentic AI boom?
😭 Meta’s embrace of AI is making its employees miserable.
🤦♀️ The AI fitness instructors selling unreal gains.
🙅♂️ TikTok scales back AI-generated video descriptions after absurd errors.
👩⚖️ DOGE used ChatGPT in a way that was both dumb and illegal, judge rules.
🛠️ Trending Tools
For the Real Estate Pros: RoomX AI transforms empty room photos with pro furniture placement in 30 seconds. With 10+ styles at just $0.02 per image, it’s 99% cheaper than traditional staging.
For the Email Marketers: EmailVerify ensures your hard work doesn't end up in the spam folder. This enterprise-grade tool scrubs your list of fake or dead addresses, boost deliverability and protect your sender reputation, ensuring your ROI stays high and your emails actually reach human eyes.
For the Solo Livers: Demumu is a lightweight safety app that gives you and your loved ones peace of mind. It uses a simple daily check-in system that automatically notifies your emergency contacts if you ever fail to check in.That way you’re never truly alone in an emergency.
For the Inbox Warriors: Shortwave – If your Gmail feels like a graveyard of unread threads, Shortwave is the resurrection. It’s an AI-powered email client that bundles related threads, drafts perfectly phrased replies in your voice, and summarizes long chains so you can hit "Inbox Zero" in minutes, not hours.
🤖 AI Workout Of The Day: Prompt for Strategic Business Finance Planning
Big goals need more than good intentions, they need a clear financial roadmap and your business finances aren’t just numbers, they’re the story of where you are and where you’re going.
Whether you're looking to scale, stabilize, or simply breathe easier each month, a clear financial roadmap helps you make smarter decisions today that pay off tomorrow.
This prompt turns AI into your virtual CFO, guiding you step-by-step through budgeting, reinvestment, debt reduction, and long-term planning, all tailored to your goals.
Here Are Some Tips to Deploy This Prompt Effectively:
Include specific numbers where possible: revenue, expenses, debts, margins, etc.
Include Your Industry: Finance strategies for SaaS companies differ from those for restaurants or agencies.
Define your business goals clearly: Are you aiming to expand, pay off debt, or boost profitability? Get specific.
Mention your risk appetite (e.g., conservative, moderate, aggressive) so the roadmap fits your comfort level.
Ask for milestone checkpoints and cash flow advice to stay realistic.
💡 Prompts to try:
Act as a strategic financial planner for small businesses. Help me create a financial roadmap to achieve the following business-specific goals: [INSERT GOALS — e.g., expanding operations, reducing debt, improving cash flow, long-term sustainability].
First, assess the current financial health of my business based on this information:
-Current revenue streams
-Monthly operational costs
-Existing debts and liabilities
-Projected growth rate
-My industry and business model
-My risk appetite (e.g., conservative, moderate, aggressive)
Then, build a step-by-step financial roadmap including:
-Budgeting recommendations
-Reinvestment strategies
-Risk mitigation tactics
-Milestone-based targets with timelines
-Tips for improving financial discipline and forecasting
Tailor the advice to my industry and business stage. Keep it practical and forward-looking.Is this your AI Workout of the Week (WoW)? Cast your vote!
That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What'd you think of today's email?
Your feedback means a lot to us and helps improve the quality of our newsletter.
