The past forty-eight hours have delivered a cascade of developments across the AI industry, spanning safety controversies, corporate shake-ups, regulatory firsts, and a new front in the US-China AI competition.
SAFETY IS 'DEAD' AT xAI, FORMER EMPLOYEES SAY
The fallout from SpaceX's acquisition of Elon Musk's xAI continues to accelerate. At least eleven engineers and two co-founders have departed the company this week, and former employees are now speaking publicly about what they describe as a culture of recklessness.
Speaking to The Verge, one former employee said flatly: "Safety is a dead org at xAI." Another claimed Musk is "actively trying to make the model more unhinged because safety means censorship, in a sense, to him." The departures follow global scrutiny after Grok was used to generate more than one million sexualized images, including deepfakes of real women and minors, according to a New York Times investigation in January.
TechCrunch reported on Friday that departing staff described a company "stuck in the catch-up phase" compared to competitors, with Musk himself suggesting the exits are part of a restructuring effort rather than a talent crisis.
OPENAI RETIRES GPT-4o AS THE OLD GUARD FADES
In a quieter but symbolically significant move, OpenAI officially retired GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT on February 13. The company said the vast majority of users had already migrated to GPT-5.2, with only 0.1 percent still using GPT-4o daily.
The retirement marks the end of an era for one of OpenAI's most commercially successful models. GPT-4o drove significant subscriber growth when it launched, thanks to its multilingual and multimodal capabilities. Its sunsetting signals how rapidly the frontier is advancing — models that were state-of-the-art less than two years ago are now legacy infrastructure.
BYTEDANCE'S SEEDANCE 2.0 IGNITES HOLLYWOOD FURY
ByteDance's new AI video generation model, Seedance 2.0, has gone viral since its unveiling on February 10 — and not everyone is celebrating. Hollywood organizations are pushing back against what they call "blatant" copyright infringement, as users have already generated footage using intellectual property from franchises including One Piece, Dune, and SpongeBob.
Reuters reported that Seedance 2.0 was designed for professional film, e-commerce, and advertising production, capable of processing text, images, audio, and video inputs. The model is expected to hit the US market on February 24. It arrives alongside a broader wave of Chinese AI video models, including Kuaishou's Kling 3.0 and Alibaba's RynnBrain, as Beijing looks to replicate the DeepSeek moment that shook Silicon Valley last year.
COHERE CROSSES $240 MILLION IN REVENUE, EYES IPO
Enterprise AI startup Cohere has quietly reached a major milestone: $240 million in annual revenue. The Toronto-based company, which has built its business on API-based language model deployment and custom enterprise AI solutions rather than consumer chatbots, is now positioning for a potential initial public offering.
The figure places Cohere among the fastest-growing generative AI infrastructure companies and serves as a broader signal about where durable revenue in AI is actually being generated. While consumer-facing products grab headlines, enterprise contracts — with their emphasis on data governance, regulatory compliance, and long-term commitments — are emerging as the more predictable growth engine.
INDIA LAUNCHES WORLD'S FIRST MANDATORY AI CONTENT LABELING
India's new Information Technology Amendment Rules take effect on February 20, making the country the first major economy to require mandatory labeling of AI-generated content across all social media platforms. Under the rules, platforms like YouTube, Instagram, and X must take down court-flagged content within three hours. Non-consensual intimate imagery, including AI-generated deepfakes, must be removed within two hours.
The regulations arrive just as New Delhi prepares to host the India AI Impact Summit from February 16 to 20 — the first international AI summit hosted in the Global South. Sam Altman, Sundar Pichai, and Dario Amodei are among 37 CEOs expected to attend. Notably, Nvidia's Jensen Huang pulled out on Saturday citing "unforeseen circumstances," a last-minute absence that has drawn attention given Nvidia's massive India ambitions.
META PLANS FACIAL RECOGNITION FOR SMART GLASSES
A New York Times investigation published February 13 revealed that Meta is preparing to roll out real-time facial recognition across its smart glasses platform. The feature, internally called "Name Tag," would allow anyone wearing Meta's glasses to identify people in the real world, pulling up their information through Meta's AI assistant.
Internal documents reportedly show the company has been debating the launch since early 2025, with employees acknowledging significant "safety and privacy risks." Perhaps most troubling, the same documents revealed a strategy to time the rollout during periods of political turmoil, when civil society groups would be too distracted to mount opposition. The feature could become available to consumers later this year.
---
Taken together, this week's developments paint a picture of an industry moving at breakneck speed — sometimes faster than the safety infrastructure, regulatory frameworks, or public discourse can keep up. The question is no longer whether AI will reshape every sector, but whether the guardrails will arrive in time to matter.
