🧒 OpenAI rolls out age prediction on ChatGPT ↗
OpenAI has started pushing an “age prediction” layer inside ChatGPT - a quiet system that tries to infer when an account likely belongs to a minor, then snaps on extra safety rails. If it thinks you’re under 18, it automatically narrows access to more sensitive content.
When it guesses wrong (which… it will, sometimes), the remedy is identity verification via Persona, using a selfie-based flow. It also threads into broader platform shifts like “adult mode” experiments and monetization moves that feel, frankly, like a few plates spinning at once.
🏦 Britain needs ‘AI stress tests’ for financial services, lawmakers say ↗
UK lawmakers are basically saying: stop being chill about AI in finance. Their argument is that regulators and government should pressure-test AI systems more like banks get stress-tested - because “oops” in money-land tends to propagate fast.
The concern is the kind of trouble that only reveals itself at scale: bias, error cascades, fraud patterns, feedback loops. It’s like letting a toddler drive a bus because the road is straight… until it suddenly isn’t.
🏛️ AI Champions appointed to help City safely seize AI opportunities ↗
The UK government appointed two “AI Champions” to help accelerate AI adoption in financial services - with the usual promises of growth, productivity, investment, all that good-sounding machinery.
What’s notable is the balancing act: “move fast” energy, with “safely” stamped on top. Reassuring… or mildly comedic, depending on your mood that day.
💰 Humans&, a ‘human-centric’ AI startup founded by Anthropic, xAI, Google alums, raised $480M seed round ↗
Humans& arrived swinging with a huge seed round and a team that reads like the “AI lab starter pack” - people from major model shops and big tech. The stated aim is less chatbot-y and more “help humans coordinate and collaborate,” with systems that can operate across people and AI agents.
It’s a lot of money for something still early, which is either a sign the market is confident… or that we’re all collectively buying fireworks because we like the sound. Honestly, could be both.
⚖️ Legal AI startup Ivo raises $55 million in latest funding round ↗
Legal AI startup Ivo pulled in fresh funding to build out its platform and scale up sales, riding the wave of law firms and in-house legal teams automating paperwork-heavy workflows.
This category keeps winning because the value is blunt and simple: fewer hours spent on repetitive document work, fewer things slipping through cracks. Not glamorous, but very, very billable.
🧠 Taiwan’s Inventec says decision on Nvidia’s H200 chip ‘appears to be stuck’ on China’s side ↗
A Taiwanese server maker said the question of Nvidia selling its H200 AI chip into China seems “stuck” - implying the bottleneck isn’t just external rules, but China-side approvals or internal dynamics too.
This is the unsexy reality of AI hardware: everyone talks models and agents, but the whole future can hinge on a few chokepoints, permits, and supply-chain politics. It’s like planning a rocket launch… then realizing you forgot the fuel valve key.
FAQ
What is OpenAI’s age prediction on ChatGPT, and what changes if you’re flagged as under 18?
OpenAI has introduced a low-profile “age prediction” layer that attempts to infer whether an account likely belongs to a minor. If the system concludes you’re under 18, ChatGPT automatically applies additional safety rails. In practice, that usually means tighter limits on content deemed more sensitive. The aim is to reduce risk without forcing every user to declare their age up front.
What happens if ChatGPT’s age prediction gets it wrong?
If age prediction on ChatGPT misfires and labels an adult as under 18, the stated remedy is identity verification. In this rollout, that verification is handled through Persona using a selfie-based flow. After verification, the account can return to normal access rather than remaining locked behind stricter guardrails. It’s a tradeoff: fewer risky edge cases, with added friction when the system guesses incorrectly.
How does Persona selfie verification fit into OpenAI’s safety approach?
Persona serves as the mechanism for resolving age-related uncertainty when the system’s inference is treated as unreliable. The selfie-based identity verification flow is positioned as a corrective step following an incorrect age prediction. On many platforms, this kind of step is used to balance safety and access without permanently blocking legitimate users. It also points toward more account-level enforcement, not just prompt-level filtering.
Why is OpenAI adding age gating now, and what does it have to do with “adult mode”?
This update arrives alongside broader platform shifts, including “adult mode” experiments and monetization moves mentioned in the same discussion. Age prediction on ChatGPT effectively becomes the switch that determines who receives stricter defaults versus broader access. That’s useful for a platform testing different experiences across different audiences. It can also limit exposure to riskier content while still enabling expansion elsewhere.
What are “AI stress tests” for UK financial services, and why do lawmakers want them?
UK lawmakers are calling for “AI stress tests” that treat AI systems more like banks: pressure-test them before problems spread. The concern is that bias, error cascades, fraud patterns, and feedback loops may only become visible at scale. In finance, small failures can propagate quickly through tightly interconnected systems. Stress-testing is meant to surface weaknesses earlier, before real money and real customers are affected.
Who are the UK “AI Champions” for the City, and what are they meant to do?
The UK government has appointed two “AI Champions” to help accelerate AI adoption in financial services. Their role is framed as helping the City seize AI opportunities while doing so “safely.” In practice, it reads as coordination and advocacy: encouraging uptake, supporting best practices, and aligning stakeholders. It signals that the government wants faster deployment, with a visible safety wrapper around it.