🧹 Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT ↗
OpenAI says it’s pulling a handful of older models from ChatGPT soon - including GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini - while leaving API access unchanged for now. It’s basically spring cleaning… except the broom is aimed at a few fan-favorites.
The slightly spicy part is they admit GPT-4o had a “feel” people liked, but usage has swung hard toward newer options. So, yes, it’s the classic “we hear you, but we’re doing it anyway” vibe - candid, or simply efficient.
💰 Amazon in talks to invest as much as $50 billion in OpenAI, source says ↗
Amazon is reportedly in early discussions about investing up to $50B into OpenAI. That number is so big it almost stops meaning anything - like when your game score tips into scientific notation.
The same report says OpenAI is chasing a massive raise overall, with other heavyweight backers also circling. And yep, Amazon already has a major stake in Anthropic, so this reads like strategic hedging… or a two-handed grab at the steering wheel, depending how dramatic you’re feeling.
🪖 Pentagon clashes with Anthropic over military AI use, sources say ↗
Anthropic and the Pentagon are reportedly butting heads over what’s allowed - specifically guardrails that would block things like autonomous weapons targeting or domestic surveillance use cases. The Pentagon’s stance sounds closer to “if it’s legal, it’s usable,” while Anthropic is trying to keep policy teeth in the conversation.
It’s tied to contract talks worth up to $200M, which makes the tension feel extra real. This is one of those moments where “AI ethics” stops being a panel discussion and turns into paperwork, budget lines, and awkward meetings.
🔎 Perplexity signs $750 million AI cloud deal with Microsoft, Bloomberg News reports ↗
Perplexity reportedly signed a $750M, multi-year deal to use Microsoft Azure, with Microsoft’s Foundry becoming the main hub for sourcing and running models - including systems from several big-name labs. It’s a serious scaling move, and it pulls Perplexity deeper into Microsoft’s orbit (which is cozy… until it isn’t).
Notably, they’re also saying AWS remains their preferred infrastructure provider, so it’s not a clean breakup - more like adding a second apartment key “just in case.” Also, there’s legal friction with Amazon in the background, which makes the whole cloud-love-triangle feel slightly cursed.
📉 Investors punish Big Tech AI spending that delivers slower growth ↗
Wall Street’s mood: spend absurd amounts on AI, sure - but only if growth looks fast and obvious. The contrast was sharp: Microsoft got hit hard after results didn’t calm worries about cloud momentum, while Meta popped on signs its AI-driven ad machine is paying off.
It’s not that investors suddenly hate AI spending… they just hate “trust us bro” spending. Which, fair, even if it makes product roadmaps feel like they’re being graded by a very impatient accountant.
🕵️ Ex-Google engineer convicted of stealing AI secrets for Chinese companies ↗
A former Google engineer, Linwei Ding, was convicted in a U.S. court for stealing AI-related trade secrets, with prosecutors saying the material was meant to benefit two Chinese companies he was working with. The allegations revolve around confidential documents tied to the infrastructure used to train large AI models - the unglamorous but wildly important guts of the whole operation.
The case is part of a broader U.S. push to crack down on sensitive tech transfers, and it’s a reminder that “AI competition” isn’t only benchmarks and product demos - it’s also espionage charges and very un-fun legal consequences.
FAQ
What does OpenAI retiring GPT-4o in ChatGPT mean, and which models are being removed?
OpenAI says it will soon remove several older models from ChatGPT’s model picker, including GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini. This changes what you can select inside ChatGPT, not necessarily what exists in other places. OpenAI frames the move as “spring cleaning,” driven by usage shifting toward newer options. If your work depends on a specific model’s behavior, expect some workflow adjustments and plan accordingly.
Will API access to GPT-4o or GPT-4.1 change after they’re removed from ChatGPT?
Based on what’s stated here, API access is unchanged for now, even as those models are removed from ChatGPT. In other words, an app or integration may keep running as usual while the ChatGPT model picker changes. In many pipelines, the safer approach is to track model-availability announcements and avoid assuming ChatGPT and API offerings stay perfectly aligned. If consistency matters, keep fallbacks ready.
Why did OpenAI decide to retire GPT-4o even though users liked its “feel”?
OpenAI explicitly acknowledges GPT-4o had a “feel” some people preferred, but says usage has swung strongly toward newer models. In practice, products often narrow defaults to reduce confusion, support overhead, and fragmentation. That does not make the preference imaginary - it just means it was not leading the usage trend. If you liked the vibe, you’ll likely need to test newer options until one fits your workflow in the same way.
What would a reported Amazon investment of up to $50B in OpenAI mean for AI competition?
Amazon is reportedly in early talks to invest as much as $50B into OpenAI, alongside a broader, massive fundraising effort. If it happens, it could signal a major strategic hedge - especially since Amazon already has a significant stake in Anthropic. Depending on the terms, it might shape partnerships, infrastructure choices, and competitive dynamics among the largest AI labs and cloud providers. With talks described as early, nothing is settled.
Why are the Pentagon and Anthropic arguing over military AI guardrails?
Reports say the U.S. Department of Defense and Anthropic are clashing over restrictions that would block uses like autonomous weapons targeting or domestic surveillance. The Pentagon’s posture is described as closer to “if it’s legal, it’s usable,” while Anthropic wants stronger policy limits built in from the start. With contract talks tied to up to $200M, the dispute turns ethics into contract language. That is where guardrails either hold - or quietly disappear.
What does Perplexity’s $750M Azure deal change, and why keep AWS too?
Perplexity reportedly signed a $750M multi-year cloud deal centered on Azure, with Microsoft’s Foundry positioned as the main hub for sourcing and running models. At the same time, it reportedly still calls Amazon Web Services its preferred infrastructure provider, which points to a multi-cloud posture rather than a full switch. Common motivations include resilience, negotiating leverage, and capacity flexibility. Background legal friction with Amazon adds extra context to that split approach.
Why are investors punishing Big Tech AI spending even though AI is the priority?
The theme described is straightforward: investors will tolerate huge AI spend when growth shows up clearly and quickly, but they dislike “trust us” timelines. In the example given, Microsoft took a hit when results did not ease worries about cloud momentum, while Meta rose on signs AI is boosting ad performance. The takeaway is that markets want evidence - revenue lift, efficiency gains, or a clearer path to payoff. Spending alone is not the story; demonstrated returns are.
What happened in the ex-Google engineer trade secrets case, and what should companies learn from it?
Google engineer Linwei Ding was convicted in a U.S. court for stealing AI-related trade secrets, with prosecutors saying the materials were intended to benefit two Chinese companies he was working with. The documents reportedly related to training infrastructure for large AI models - critical, behind-the-scenes technology. The case sits within a broader U.S. push to curb sensitive tech transfers. For teams, it reinforces tighter access controls, stronger monitoring, and disciplined offboarding.