AI Industry Shakeup: Safety Exodus Hits OpenAI, Anthropic & xAI as $20B+ Funding Rounds Close | February 12, 2026

Daily AI Blog

February 12, 2026

📋 Quick Takeaways

Unprecedented AI safety exodus: High-profile departures hit OpenAI, Anthropic, and xAI within days — researchers warn companies are “moving too fast”

OpenAI fires VP who opposed ChatGPT “adult mode”; researcher Zoë Hitzig quits over advertising in NYT op-ed

Half of xAI’s 12 co-founders have now left Elon Musk’s AI company amid Grok deepfake controversy and reorganization

VCs break taboo: Sequoia and Altimeter invest in both OpenAI and Anthropic simultaneously as $20B+ rounds close

DeepSeek V4 expected mid-February with 1T parameters, reportedly outperforming Claude and GPT on coding benchmarks

Trump administration plans chip tariff exemptions for Amazon, Google, and Microsoft tied to TSMC’s $165B US investment

Microsoft explores superconducting power cables to solve AI data center energy bottleneck

MiniMax M2.5 launches claiming parity with Claude Opus 4.6 in coding and agentic tasks

🛡️ AI Safety & Talent Exodus

Wave of AI Safety Departures Rocks the Industry

An extraordinary cluster of high-profile resignations and firings has swept across the AI industry within a matter of days, raising urgent questions about the tension between rapid commercialization and responsible AI development.

CNN reported that in just the past few days, a number of high-profile AI staffers have decided to call it quits, with some explicitly warning that the companies they worked for are moving too fast and downplaying the technology’s shortcomings. The departures span three of the industry’s most prominent labs — OpenAI, Anthropic, and xAI — and mark a significant escalation in the ongoing struggle between safety-minded researchers and revenue-focused executives.

Why It Matters: This isn’t isolated churn. The simultaneous timing across multiple frontier labs suggests a systemic breaking point — safety researchers increasingly feel sidelined as commercial pressures mount ahead of planned IPOs for both OpenAI and Anthropic.

OpenAI Fires VP Who Opposed ChatGPT “Adult Mode”

OpenAI has fired one of its top safety and policy leaders, Ryan Beiermeister, who served as VP of Product Policy. According to The Wall Street Journal (reported Feb 10–11), Beiermeister was terminated in January after a male colleague accused her of sex discrimination — an allegation she has firmly denied.

Key Details:

Beiermeister had voiced opposition to OpenAI’s planned “adult mode” for ChatGPT, which would allow erotica for verified adults
An internal advisory council on “well-being and AI” also urged leadership to reconsider the feature
OpenAI stated her departure was “not related to any issue she raised while working at the company”
CEO Sam Altman has defended the feature as part of a “treat adult users like adults” philosophy, with launch planned in Q1 2026
OpenAI’s IPO odds on Polymarket dropped from 60% to 47% within 24 hours of the report

Industry Context: The firing follows a pattern of safety-focused executives clashing with commercial ambitions at frontier labs. OpenAI projects a $14 billion loss in 2026, creating pressure to explore new revenue sources including adult content.

Source: TechCrunch | Newsweek | TechSpot | Benzinga

OpenAI Researcher Zoë Hitzig Publishes Resignation in New York Times Over ChatGPT Ads

On February 11, 2026, Zoë Hitzig — a researcher at OpenAI for the past two years — published a resignation op-ed in The New York Times citing deep reservations about OpenAI’s emerging advertising strategy.

Key Concerns Raised:

ChatGPT’s archive of deeply personal user data — including medical fears, relationship problems, and beliefs about God — creates unprecedented manipulation potential when combined with advertising
Hitzig warned that OpenAI is building an “economic engine” that will create incentives to override its own principles around user privacy
She wrote: “I once believed I could help the people building A.I. get ahead of the problems it would create. This week confirmed my slow realization that OpenAI seems to have stopped asking the questions I’d joined to help answer.”

Why It Matters: Hitzig’s public resignation underscores growing concern that as AI chatbots become deeply personal tools, their monetization through advertising represents a fundamentally different — and potentially dangerous — model compared to traditional search or social media advertising.

Source: National Today | CNN

Half of xAI’s Founding Team Has Now Left the Company

Elon Musk’s xAI is facing a significant leadership crisis. Co-founders Yuhuai (Tony) Wu and Jimmy Ba resigned within 24 hours of each other (Feb 9–10), meaning exactly half of the original 12-person founding team has now departed.

Departures Timeline (Recent):

Feb 9: Tony Wu, co-founder and reasoning lead — “It’s time for my next chapter”
Feb 10: Jimmy Ba, co-founder and research/safety lead — “Recursive self improvement loops likely go live in the next 12 months. It’s time to recalibrate my gradient”
Feb 10: Vahid Kazemi, ML PhD — “All AI labs are building the exact same thing, and it’s boring”
Feb 7: Shayan Salehian, product infrastructure — left to “start something new”
Feb 6: Ayush Jaiswal, engineer — “taking a few months to spend time with family”

Previously departed co-founders: Igor Babuschkin, Kyle Kosic, Christian Szegedy, and Greg Yang (stepped back due to Lyme disease)

Musk posted on X that xAI was “reorganized” to speed up growth, which “unfortunately required parting ways with some people.” However, the exodus comes as the company faces global backlash and regulatory probes over Grok’s ability to generate non-consensual deepfake images, including of children.

Why It Matters: With an IPO looming and ambitious plans for orbital data centers, xAI’s ability to retain top-tier talent is being tested. At least three departing engineers are starting something new together, which could signal the birth of a new competitor.

Source: TechCrunch | CNBC | TechCrunch

Anthropic Safety Lead Mrinank Sharma Resigns, Warns “The World Is in Peril”

Mrinank Sharma, who led Anthropic’s Safeguards Research team, announced his resignation on February 9 in a public letter that quickly garnered over a million views on X.

Key Quotes from His Letter:

“The world is in peril. And not just from AI, or bioweapons, but from a whole series of interconnected crises unfolding in this very moment”
“Throughout my time here, I’ve repeatedly seen how hard it is to truly let our values govern our actions… where we constantly face pressures to set aside what matters most”

Context: Sharma holds a PhD in machine learning from Oxford. His work at Anthropic included developing defenses against AI-assisted bioterrorism and researching AI sycophancy. A study he published last week showed that AI chatbot interactions producing distorted perceptions of reality “occur daily.” He plans to pursue a poetry degree and “devote myself to the practice of courageous speech.”

Anthropic stated it was grateful for Sharma’s work but noted he was not the head of safety nor in charge of broader safeguards. Other researchers including Harsh Mehta and Behnam Neyshabur have also recently left Anthropic.

Source: eWeek | American Bazaar | MediaPost

💰 AI Funding & Investment

VCs Break Taboo by Backing Both OpenAI and Anthropic Simultaneously

In a striking departure from traditional venture capital norms, Bloomberg reported on February 11 that top-tier VCs are now investing in both OpenAI and Anthropic — direct competitors racing to close two of the largest funding rounds in history.

Key Details:

Sequoia Capital has invested in multiple OpenAI rounds since 2021 and is now expected to invest in Anthropic’s $20B+ round
Altimeter Capital, which has called OpenAI its biggest bet ever, is set to invest more than $200 million in Anthropic
Anthropic’s round is expected to bring in more than $20 billion at a $350 billion valuation, with Nvidia and Microsoft contributing up to $15 billion combined
OpenAI is simultaneously assembling a separate $100 billion fundraising round

Why It Matters: The practice of backing direct competitors was once considered taboo in startup investing. Its normalization signals that investors view the AI frontier as a multi-winner market — or that the capital requirements are so enormous that companies can no longer afford exclusivity from their investors. Both companies are also preparing potential IPOs.

Source: Bloomberg | TechCrunch

AI Capex Frenzy: $660 Billion Planned for 2026, Hyperscaler Guidance Up 24%

The AI infrastructure buildout continues to accelerate at a staggering pace. According to Wells Fargo analysis reported by Fortune on February 10–11, hyperscaler capex guidance for 2026 is up 24% — $117 billion more than last year.

Key Figures:

Total planned AI infrastructure spend: $660 billion in 2026, $1.3 trillion through 2027
Meta: Up to $185 billion in capex for 2026
Amazon: Up to $200 billion planned
Tesla: $20 billion earmarked for factories and AI compute infrastructure
Companies are increasingly taking on debt to fund the buildout, with Bank of America forecasting ~$140 billion in hyperscaler direct debt issuance for 2026

Market Context: 75% of S&P 500 companies have reported Q4 earnings, with earnings per share 12% higher year-over-year and 5% above consensus estimates — suggesting the real economy is keeping pace with AI investment narratives.

Source: Fortune

🔧 AI Models & Technology

DeepSeek V4 Imminent — Markets Brace for Potential Repeat Shock

DeepSeek’s V4 model is expected to launch around February 17, 2026, coinciding with Lunar New Year — mirroring the timing of DeepSeek R1’s release, which triggered a $1 trillion tech stock selloff in January 2025.

What We Know:

Architecture: Incorporates Manifold-Constrained Hyper-Connections (mHC) and Engram conditional memory for near-infinite context retrieval
Scale: Reportedly 1 trillion parameters with 1M+ token context window
Performance: Internal benchmarks claim V4 outperforms Claude and GPT series in complex coding and long-context tasks
Efficiency: DeepSeek Sparse Attention reduces computational costs by ~50%
Accessibility: Designed to run on consumer hardware — dual RTX 4090s or single RTX 5090
Open-source: Expected to follow DeepSeek’s tradition of open-weight releases

Market Impact: Investors are bracing for potential disruption. If V4 demonstrates comparable or superior performance to US frontier models at dramatically lower cost, it could once again raise questions about whether massive AI capex spending is justified.

Source: The Motley Fool | The Information | Introl

MiniMax M2.5 Launches, Claims Parity with Claude Opus 4.6 in Coding

On February 12, 2026, Chinese AI startup MiniMax officially released its flagship model M2.5, billed as “the world’s first production-level model designed natively for Agent scenarios.”

Key Claims:

Full-stack programming support across PC, App, and cross-platform development
Industry-leading (SOTA) performance in office productivity tasks including Excel, research, and presentation creation
Comprehensive performance directly competing with Claude Opus 4.6
Company’s market valuation reportedly surged past 180 billion yuan (~$25B)

Why It Matters: MiniMax’s release adds to the growing competitive pressure from Chinese AI labs. Combined with the imminent DeepSeek V4, it underscores that the performance gap between Chinese and Western AI models continues to narrow rapidly.

Source: AIBase

🏛️ AI Policy & Infrastructure

Trump Administration Plans Chip Tariff Exemptions for Big Tech Amid AI Data Center Push

The Trump administration is preparing to exempt Amazon, Google, and Microsoft from upcoming semiconductor tariffs, according to the Financial Times (reported Feb 10–11). The exemptions are closely tied to investment commitments from TSMC.

Key Details:

TSMC has pledged $165 billion in U.S. semiconductor manufacturing investments
Under the proposed plan, TSMC could grant tariff exemptions for its US clients
Taiwanese firms building US plants could import up to 2.5x planned capacity tariff-free during build-out
The Commerce Department is preparing the carve-outs, though final approval by President Trump is still pending

Strategic Implications: The tariff relief creates a two-tier system — hyperscalers with direct TSMC relationships get preferential access, while smaller AI infrastructure providers and independent chip buyers face full tariff costs. This could further consolidate compute power among the largest cloud providers.

Source: TrendForce | EconoTimes | Windows Report

Microsoft Explores Superconducting Power Cables for AI Data Centers

Microsoft announced on February 10–11 that it is exploring high-temperature superconductor (HTS) cables for data center power systems — a potential breakthrough for the AI infrastructure bottleneck.

Technical Details:

HTS cables deliver the same electricity as traditional cables in a dramatically smaller footprint
Uses ceramic-like material that transports electricity with zero resistance, eliminating energy loss
Cables can be more than 10x smaller and lighter than traditional copper cables
Microsoft has invested in HTS startup VEIR, which recently demonstrated a 3-megawatt data center power delivery system

Why It Matters: The AI industry’s data centers are being built faster than power grids can be upgraded. HTS technology could enable Microsoft to increase electrical density without expanding physical infrastructure — potentially shortening deployment timelines from years to months. However, the technology remains in development stage and key materials (rare-earth barium copper oxide tape) are largely produced in China.

Source: Reuters/Yahoo Finance | Windows Central

🏢 Corporate AI Strategy

Salesforce Lays Off ~1,000 Employees Amid AI Restructuring

Salesforce has quietly laid off nearly 1,000 employees in February 2026, primarily impacting its Agentforce AI and marketing teams, accompanied by a significant executive shakeup. Workday simultaneously announced 400 job cuts in its own AI-driven strategic shift.

Industry Context: The layoffs come in the wake of a broader SaaS sector selloff triggered by Anthropic’s Claude Cowork features and similar AI-native tools that threaten traditional enterprise software business models. The market is increasingly questioning whether established SaaS companies can defend their positions against AI platforms that own workflows directly rather than selling point solutions.

Source: Storyboard18

Major AI Companies Launch European Accelerator at Station F

Wired reported that Meta, Microsoft, Google, Anthropic, OpenAI, and Mistral are collaborating on a European accelerator program based at Station F in Paris.

Program Details:

Offers significant compute credits and partnerships to European startups
Aims to help startups build on the major AI players’ models and tooling
Part of a broader effort to keep European AI talent from relocating and accelerate commercialization

Strategic Signal: As model performance differences narrow, the real competitive moat shifts to developer loyalty — credits, integrations, distribution, and becoming the “default stack” that startups choose early. This accelerator represents a new front in the platform war, moving from pure model capability to ecosystem building.

Source: Wired | Tech Startups

ByteDance Negotiates with Samsung for Custom AI Chip Production

ByteDance is in negotiations with Samsung to manufacture custom AI chips, seeking to diversify its supply chain and reduce dependence on external providers like NVIDIA.

Strategic Context:

ByteDance plans to increase AI infrastructure spending to $23 billion
Custom chip production with Samsung would provide a non-US fabrication path
The deal would represent a significant shift in AI chip supply chain dynamics, with a major Chinese tech company securing dedicated capacity from a Korean chipmaker

Why It Matters: As US chip export restrictions tighten and tariffs reshape the semiconductor landscape, Chinese tech giants are actively seeking alternative paths to AI compute — a trend that could accelerate the fragmentation of the global AI hardware ecosystem.

Source: LatestLY

📊 Market Impact Analysis

The convergence of safety departures, record funding, and technology disruption signals an AI industry approaching a critical inflection point. Key observations:

Safety vs. Speed Collision: The simultaneous departures across OpenAI, Anthropic, and xAI represent the most significant wave of AI safety talent exodus to date. As these companies race toward IPOs, the commercial pressure to ship products faster appears to be overwhelming internal safety voices.

Capital Concentration Intensifies: With $120B+ being raised between OpenAI and Anthropic alone, and VCs abandoning exclusivity norms to invest in both, the AI funding landscape is entering unprecedented territory. The question is no longer whether there’s enough capital — it’s whether the returns can justify the concentration.

Chinese AI Challenge Grows: DeepSeek V4 and MiniMax M2.5 represent the latest evidence that Chinese AI labs are closing the gap with Western frontier models. The approaching DeepSeek V4 launch could be the most consequential AI model release of 2026, with potential to once again reshape market assumptions about the economics of frontier AI.

Infrastructure Meets Reality: Microsoft’s superconductor exploration, TSMC tariff carve-outs, and ByteDance’s Samsung chip negotiations all reflect the same underlying tension — AI’s growth is increasingly constrained not by algorithms but by atoms: power, chips, and physical infrastructure.

🔮 Looking Ahead

Key Events to Watch:

DeepSeek V4 launch (~Feb 17) — potential market-moving event
Anthropic $20B+ funding round closing — could set new private company valuation records
OpenAI “adult mode” launch in Q1 2026 — will face intense scrutiny after executive departures
NVIDIA earnings (Feb 25) — critical test for AI infrastructure thesis
India AI Impact Summit (Feb 16–20) — Nvidia, Google, OpenAI among 400 exhibitors

Stay Updated: Follow us for comprehensive daily AI news, policy analysis, and investment tracking.

Last Updated: February 12, 2026, 10:30 PM CST