#011: OpenAI Gets Honest, TikTok Gets Mindful & SoundCloud Listens

KIDS ONLINE SAFETY ACT RETURNS – BIG TECH IS SPLIT, CRITICS AREN’T SOLD

The bipartisan Kids Online Safety Act (KOSA) has been reintroduced in Congress, and if passed, it would mark the biggest U.S. internet safety overhaul since the 1998 COPPA law. The bill would hold social platforms legally responsible if they don’t protect minors from harms like eating disorders, sexual exploitation, and self-harm(¹). KOSA sailed through the Senate last year but then stalled in the House.

Critics, from the ACLU to the EFF, warn that even a well-intentioned “duty of care” could morph into censorship and surveillance. Yet KOSA has gained heavyweight support: Microsoft, Snap, and even X (Twitter) helped shape the new draft.

Apple just joined the fight, calling the bill a “meaningful impact on children’s online safety.” Major players like Google and Meta remain opposed. Still, with Big Tech divided and political momentum high, the stage is set for a renewed battle over what it means to keep kids safe online.

CANCELING SUBSCRIPTIONS REMAINS A HASSLE AS FTC DELAYS CLICK-TO-CANCEL RULE

The Federal Trade Commission (FTC) has postponed the enforcement of its new “click-to-cancel” rule by 60 days, moving the effective date to July 14, 2025(²). The regulation, part of the Negative Option Rule, mandates that businesses make it as straightforward to cancel subscriptions as it is to sign up for them.

The rule addresses widespread consumer frustration with convoluted cancellation processes, often involving navigating through multiple steps or contacting customer service. While the FTC cites the complexity of compliance as the reason for the delay, critics argue that businesses have already had ample time to prepare, and the postponement benefits companies that make canceling subscriptions intentionally difficult.

Other News

• 💣 OpenAI’s Bunker Mentality
  In a revealing moment, OpenAI co-founder Ilya Sutskever told new researchers, “We’re definitely going to build a bunker before we release AGI.” The comment underscores internal tensions at OpenAI over the potential risks of artificial general intelligence and the company's shift from its nonprofit roots to a profit-driven model (⁴).

• 🏫 Schools Caught Off Guard by ChatGPT
  Public records reveal that many U.S. school districts were unprepared for the impact of ChatGPT, with some educators receiving training from pro-AI consultants while others had no guidance (⁵). The documents highlight a patchwork response to AI in education, leaving schools scrambling to address challenges like cheating and curriculum changes.

• 🔍 Flock’s Surveillance Tool Raises Privacy Concerns
  Leaked materials show that Flock, known for its license plate readers, is developing a tool called Nova that combines vehicle data with personal information from various sources(⁶). The system aims to help law enforcement “jump from LPR to person,” raising alarms about potential privacy violations and the use of data without warrants.

#005: CAN AI COMPANION BOTS REPLACE REAL FRIENDS? THE FRIENDSHIP EXPERT WEIGHS IN

In this episode, we cover:

• Why AI companion bots can’t replace real friends

• How social media is quietly complicating friendship dynamics and fueling comparison

• The risks of sharing friendship drama online (and why it might backfire)

• How friendships can breed “digital surveillance” habits with read receipts and location tracking

If you’re wondering how social media might have impacted your friendships or if you know someone who is trying out a “companion bot,” Danielle has some things to share.

Stay up-to-date with AI

The Rundown is the most trusted AI newsletter in the world, with 1,000,000+ readers and exclusive interviews with AI leaders like Mark Zuckerberg, Demis Hassibis, Mustafa Suleyman, and more.

Their expert research team spends all day learning what’s new in AI and talking with industry experts, then distills the most important developments into one free email every morning.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

WHEN BIG TECH LISTENS: SMALL WINS FROM OPENAI, TIKTOK & SOUNDCLOUD

This week’s hopeful news comes from the companies we usually monitor closely. We’re giving credit where credit is due to these three tech companies, which are seemingly trying to right their previous wrongs.

We’re here for it!

For users under 18, the feature is enabled by default. If a teen is on TikTok past 10 p.m., their feed is interrupted with a calming breathing exercise and soft music to nudge them to log off. If they ignore the prompt and keep scrolling, a second full-screen alert will tell them to go to bed, and adults can opt in to these bedtime reminders via the app’s settings.

It’s part of TikTok’s response to growing concerns about youth screen time and mental health: the company is even donating $2.3 million in ad credits to mental health organizations across 19 countries as part of this wellness initiative.

PHOTO: Eugene Gologursky

OpenAI says it will update these safety metrics with each major model update and share progress as it develops better ways to measure and improve AI safety. This transparency push comes after months of criticism that OpenAI was too secretive and rushed to deploy AI updates.

We hope that by voluntarily spotlighting its models’ flaws and fixes, OpenAI might nudge the rest of the industry toward a safer, more accountable AI ecosystem.

SOUNDCLOUD PULLS THE PLUG ON AI AUDIO TRAINING

In last week’s newsletter, we shared how SoundCloud got caught in the AI backlash after quietly tweaking its terms to allow AI training on user content(⁹). This week, the company hit rewind, admitting the update was “too broad” and officially banning the use of music or voice for generative AI models. This is a win for creators and a sign that pushback still works.

If you enjoyed this newsletter, please support Log Out by subscribing to our podcast or sharing this newsletter with a friend.