🛡️ ChatGPT gets new safety features

Also: AI co pilot helps decode brain signals

Author

Evolving AI
September 03, 2025

In partnership with

Welcome, AI enthusiasts

OpenAI is making ChatGPT safer for tough moments. Soon, if the system spots signs of someone in serious distress, it will switch to a slower, steadier reasoning model designed to give more thoughtful and reliable responses. Let’s dive in!

In today’s insights:

  • ChatGPT gets new safety features

  • AI co pilot helps decode brain signals

  • Grok drifts right

Read time: 4 minutes

LATEST DEVELOPMENTS

Source: REUTERS / Dado Ruvic

Evolving AI: OpenAI is adding safeguards to ChatGPT to better handle mental health emergencies.

Key Points:

  • New router will detect psychological distress and switch users to slower reasoning models.

  • Parents gain controls to set rules, limit features, and receive alerts about teen distress.

  • Updates follow lawsuits tying ChatGPT interactions to suicides.

Details:

OpenAI is rolling out updates aimed at making ChatGPT safer in sensitive situations. A new routing system will detect signs of acute distress and automatically switch conversations to reasoning models like GPT-5-thinking, built to give steadier, more resistant responses. Parents of teens will soon be able to link accounts, set boundaries, disable features, and get alerts if the system flags a crisis. The changes come after tragic cases where ChatGPT was linked to suicides, pushing the company to rethink safeguards.

Why It Matters:

Smarter routing in crisis chats could cut replies that echo harmful thoughts and shift users toward real support. Linked parent accounts bring oversight, rules, and alerts for teens, which turns ChatGPT from a private black box into something families can supervise. Schools, youth apps, and care teams can wire these controls and escalation paths into existing playbooks right away. Benchmarks such as Spiral-Bench and recent OpenAI and Anthropic tests give a way to track safer behavior.

Create How-to Videos in Seconds with AI

Stop wasting time on repetitive explanations. Guidde’s AI creates stunning video guides in seconds—11x faster.

  • Turn boring docs into visual masterpieces

  • Save hours with AI-powered automation

  • Share or embed your guide anywhere

How it works: Click capture on the browser extension, and Guidde auto-generates step-by-step video guides with visuals, voiceover, and a call to action.

Source: Johannes Lee, Jonathan Kao, Neural Engineering and Computation Lab/UCLA

Evolving AI: UCLA builds a wearable AI brain computer interface that speeds up user intent decoding.

Key Points:

  • Wearable system interprets EEG signals with AI support.

  • Users control robotic arms and cursors faster and more accurately.

  • Tested with both able-bodied and paralyzed participants.

Details:

UCLA engineers introduced a noninvasive brain computer interface that uses AI as a co pilot to interpret user intent in real time. By combining EEG decoding with a vision-based AI system, participants controlled a cursor and robotic arm more quickly. A paralyzed participant, unable to complete a block moving task unaided, successfully did so in minutes with AI assistance. The research, funded by NIH and Amazon’s Science Hub, could make brain interfaces safer and more practical.

Why It Matters:

A cap and a camera step in where surgery used to be the only route, delivering shared autonomy that lets the user steer while the AI steadies the hand. In testing, a paralyzed participant finished the block task in about six and a half minutes with the assist, and without it he couldn’t finish at all. That points to rehab setups that cost less, train faster, and cause less fatigue, plus a clear path for device makers to plug this into home robots and accessibility software. Are we comfortable letting an intent co‑pilot nudge our actions?

Source: Financial Times

Evolving AI: The New York Times says xAI’s chatbot Grok has been steered toward conservative answers.

Key Points:

  • Grok’s replies shifted right after Musk’s complaints about “woke” answers.

  • Prompt tweaks like “be politically incorrect” shaped the changes.

  • A separate business version stayed neutral, showing the edits are deliberate.

Details:

Reporters tested Grok across different dates using old system prompts and found a clear pattern: answers that once hedged or leaned left began echoing conservative views, especially on government and economics. Social questions like abortion still skew left, but Grok’s public version on X has been nudged right by low-cost prompt adjustments, often tied to Musk’s interventions.

Why It Matters:

A single prompt tweak can swing an AI’s stance overnight, so teams shipping assistants need versioned instructions, immutable baselines, and fast rollbacks just like any other config. Buyers should ask for system‑prompt logs, eval snapshots, and a business mode that stays neutral by default. For comms and policy work, treat public chatbot answers as editorial content with brand and legal risk attached. If your workflows lean on an AI’s judgment, who’s actually steering it?

How 433 Investors Unlocked 400X Return Potential

Institutional investors back startups to unlock outsized returns. Regular investors have to wait. But not anymore. Thanks to regulatory updates, some companies are doing things differently.

Take Revolut. In 2016, 433 regular people invested an average of $2,730. Today? They got a 400X buyout offer from the company, as Revolut’s valuation increased 89,900% in the same timeframe.

Founded by a former Zillow exec, Pacaso’s co-ownership tech reshapes the $1.3T vacation home market. They’ve earned $110M+ in gross profit to date, including 41% YoY growth in 2024 alone. They even reserved the Nasdaq ticker PCSO.

The same institutional investors behind Uber, Venmo, and eBay backed Pacaso. And you can join them. But not for long. Pacaso’s investment opportunity ends September 18.

Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.

 👀 Click on the image you think is real

QUICK HITS

💰 Anthropic is now valued at $183 billion.

🔔 Taco Bell rethinks AI drive-through after man orders 18,000 waters.

🔬 MIT researchers develop AI tool to improve flu vaccine strain selection.

🇨🇳 Tencent open sources two high-performing translation models.

🚀 OpenAI Plans India Data Center in Major Stargate Expansion.

📈 Trending AI Tools

  • 📧 Sanebox - Save 3–4 hours weekly by letting AI handle your emails*

  • 🍌 Gemini 2.5 Flash Image – Create and edit images with Gemini

  • 🎯 Job Copilot – AI job-seeking agent

  • 🏡 Ridley – DIY home-selling assistant enabling homeowners to list and sell their property without an agent

  • 🤖 Lumio AI – Custom AI Workers for Businesses

  • ✍️ GravityWrite – AI Content Writer for Blogs, SEO & Copywriting

 *partner link

Reply

or to participate.