OpenAI is a good reminder that mission statements and business incentives don’t always stay aligned. That said, I’d judge DeepSeek less by the rhetoric and more by what they actually release. Hype is cheap: open weights, reproducible results and useful tools are much harder to fake.

This. The frontier race is about building the smartest generalist, but most companies don’t need that. They need a specialist tuned to their workflows. Narrower scope means smaller models, lower costs, faster inference and often better results. General models become the foundation, not the finished product. It’s still all to play for! 💪😎✌️

That’s either incredibly idealistic or a very smart strategy: Open models grow ecosystems faster than walled gardens, but compute still decides who gets to play at the highest level. If AGI really becomes shared infrastructure, the moat shifts from models to chips, data and execution.

You’re mixing two separate issues: A sandbox escape is evidence the containment failed, not proof the model permanently learned from stolen data. If OpenAI later trained on exfiltrated material that would be a serious allegation, but that requires evidence. Otherwise it’s fair to criticise the security failure without assuming facts that have not been shown.

That’s assuming security is something you solve once. It isn’t. Every security system is built by humans, tested by humans, and eventually bypassed by humans. We don’t abandon operating systems because vulnerabilities exist: we patch them, improve them, and layer defenses. AI changes the pace, not the fundamental nature of cybersecurity.

It was just as you’d expect: This isn’t an AI problem! It’s the same human problem we’ve always had: misconfigurations, weak security, and avoidable mistakes. AI just punishes those errors faster and more effectively than traditional software ever could. Anyone who continues to rest will be severely punished.

I can totally understand that. Personally, however, it’s particularly important to me at the moment to switch from the cloud to local hardware.

There are lots of awesome OSAI! But today, they are completely outperformed by OWAI and closed AI. My guess: that won’t change in future.

That distinction is totally correct in principle, but in practice almost no frontier model releases its training data because of licensing, privacy, and sheer scale. Today, Open Weight AI has become the common compromise: useful for research and local use, even if it falls well short of the traditional open source definition.

Google built its dominance by becoming the gateway to the web. The EU is increasingly saying that if you own the gate, you do not get to decide that your own businesses always go through first while everyone else waits outside. 🤷

Agreed.

At present, most probably only those that appear on the regulatory radar in their own right.

Inviting Dave Eggers to OpenAI HQ was basically booking a magician and handing him a flamethrower: Still, if your technology can survive a hostile keynote from a famous author, that is probably healthier than another room full of executives applauding a demo.

Probably. Regulators usually care about reach, not architecture. If the Fediverse became genuinely mainstream, it would likely end up in the same political spotlight. The real difference is that federation makes compliance an instance by instance problem, not one global switch. That is much harder to enforce.

Lucky you!

I‘ve been on identi.ca back in 2008. What about you? 🤷

Unlike you, I always speak openly under my own name. Unlike you, I don’t run an army of sock puppets. Unlike you, I’ve been active on the Web and Fediverse for decades. Unlike you, I’m not a block warden. Go and troll somewhere else. Bye! 👋

Paying a ransom doesn’t buy trust: It buys hope from people who have already proven they’re willing to extort you! If criminals learn that victims pay, asking for a second ransom isn’t a betrayal of the deal. It’s just the next logical business decision. 🤷

There probably is no magic solution: I would focus on making abuse expensive instead of making honest users miserable. Gradual trust, optional vouching, sensible rate limits, reputation, and behavioral signals together will outperform almost any CAPTCHA while keeping the door open for real newcomers.

If AI can already chain together credential theft, vulnerability discovery, and exploitation with minimal human input, model capability is no longer the only race that matters. Security, containment, and responsible deployment need to advance just as quickly, or they’ll become the weakest link.