THE RUNDOWN

The timing couldn't be worse for the AI chatbot world. Character.AI, which built its entire identity around deeply personal AI companion relationships, is heading toward bankruptcy — and New York's Attorney General has launched a multi-state investigation demanding user data protections before the doors close. If you or someone you know poured private conversations into that platform, this is a genuine concern. The question of what happens to intimate data when an AI startup dies is one the security and privacy conversation hasn't fully answered yet.

McKinsey is now asking some entry-level candidates to spend 10 minutes sparring with an AI simulation of a client as part of their hiring process. It's a clever stress test — but it also raises real questions about whether AI interviewers can fairly assess candidates from different cultural or linguistic backgrounds. Watch this space; if McKinsey's doing it, half of corporate Africa will follow within two years.

Last week it felt like Anthropic and OpenAI had pulled comfortably ahead of everyone else. This week, rivals are clawing back. Several players are shipping new model updates and capability jumps that close the gap faster than analysts expected. The AI development race is genuinely open again — which is either exciting or exhausting depending on how you're tracking it.

Startup Ricursive Intelligence is building AI systems that design semiconductor chips, learn from each iteration, and improve the next generation automatically. It's a direct swing at the compute bottleneck that's throttling AI development globally — the idea being that if humans can't design chips fast enough, let AI do it. Early results are apparently compelling enough that serious investors are paying attention.

You know those fake parking ticket texts and bogus package delivery alerts? Microsoft, working with Japanese authorities, has dismantled a cybercrime network responsible for over a billion of them. It's one of the largest coordinated takedowns of a scam-SMS operation on record, and it's a reminder that the threat isn't just sophisticated nation-state hacking — it's also industrialised nuisance crime at massive scale.

This one reads like a thriller plot, but it happened. A hack targeting internet-connected refrigeration systems hit around 700 supermarkets, costing millions and — critically — sparking actual fires. The culprit was outdated devices left exposed directly to the internet with poor security hygiene. It's a stark reminder that 'smart' infrastructure without proper security isn't smart at all.

YouTube is rolling out 'Made with AI' labels on tracks that were fully or significantly generated by artificial intelligence — similar to the disclosure labels it already uses for realistic synthetic content in videos. Streaming platform Deezer went a step further, announcing it'll let users filter out AI music entirely. The backlash from human musicians has clearly landed, and platforms are starting to respond with actual tools rather than just statements.

At TechCrunch Disrupt, Flock Safety CEO Garrett Langley defended his company's camera and microphone surveillance network against mounting criticism from the ACLU and privacy advocates. He claimed the microphones detect 'just gunshots' — but the company's own website suggests the capability is broader than that. The tension between crime-fighting utility and civil liberties is real, and it's a conversation that matters deeply for African cities being pitched similar 'smart city' security infrastructure.

Mukesh Ambani's Jio is seeking to raise $3.2 billion in what could become India's largest IPO ever. Most of the proceeds are earmarked for debt repayment as Jio pushes beyond its telecom roots into broader digital services. For African telcos watching emerging market playbooks, Jio's trajectory from pure connectivity to digital ecosystem is worth studying closely.

On a more hopeful note: NASA confirmed it's staffing up and accelerating its first Venus missions in a generation. The agency is also solving genuinely hard engineering problems, like building batteries that can survive the Moon's extreme temperature swings where standard lithium-ion simply fails. It's slow, expensive work — but it's moving again.

Today's throughline is accountability catching up with speed. From OpenAI in court to hacked fridges to surveillance CEOs on the back foot, the tech world is learning that shipping fast doesn't mean escaping consequences. Stay sharp out there.