In one of the latest developments in the AI domain, Anthropic pulled two frontier models offline under a federal export control order, then brought them back while simultaneously launching a brand new flagship. Alphabet shed hundreds of billions in market value after a wave of research talent walked out the door. Regulators on two continents moved in opposite directions, with Brussels tightening its grip on Big Tech while several US state legislatures raced to pass child safety rules for AI chatbots before their summer recesses.
This was also a week where the money kept moving even as the headlines focused on governance. Global startups raised a record $510 billion in the first half of 2026, according to Crunchbase data, and AI absorbed the overwhelming majority of that capital. Late-stage rounds landed at Together AI, TwelveLabs, and Quantum Systems, while ElevenLabs explored a stock sale that would value the voice AI company at $22 billion. Healthcare policy quietly shifted, too, with a new Medicare payment model and a tough new state AI law both taking effect on the same day.
For developers tracking model access, investors watching capital concentration, and healthcare and policy professionals monitoring compliance deadlines, this roundup pulls together the most significant, verified AI developments of the week. Here is what happened and why it matters.
This week’s top AI stories:
- Anthropic restores Claude Fable 5 and Mythos 5 after an 18-day export control shutdown
- Claude Sonnet 5 launches as the new default model; California signs a statewide Anthropic deal
- OpenAI delays the public rollout of GPT-5.6 Sol after a US government request
- Alphabet loses roughly $269 billion in market value amid a DeepMind talent exodus
- Together AI, TwelveLabs, Quantum Systems, and ElevenLabs headline a record funding week
- CMS launches its ACCESS Model as Colorado’s AI Act takes effect for healthcare providers
- Google loses its final EU appeal over a 4.1 billion euro Android antitrust fine
- State lawmakers advance AI chatbot safety bills in New York, New Jersey, and California
- CISA adds a critical LiteLLM vulnerability to its Known Exploited Vulnerabilities catalog
- Salesforce data shows enterprise agentic AI adoption nearly doubling in a year
- Grok 4.5 enters private beta as a Chinese open source model narrows the performance gap
- Palantir’s Alex Karp criticizes Frontier AI pricing as Nvidia’s Nemotron gains government traction
Anthropic Restores Claude Fable 5 and Mythos 5 After 18-Day Export Control Shutdown
Anthropic’s Mythos-class models spent most of this reporting window locked behind a federal export control order before returning to users on July 1, 2026. Claude Fable 5, first released on June 9, was pulled offline for 18 days after the US Department of Commerce issued an emergency export control directive on June 12 covering both Fable 5 and its sibling model, Claude Mythos 5. Anthropic confirmed on X that the Commerce Department had lifted the controls and that restoration would begin the following day.
The commercial terms of the comeback differ meaningfully from the original launch. Fable 5 is included within 50 percent of weekly usage limits for Pro, Max, Team, and select Enterprise plans only through July 7, after which access shifts to a separate usage credit system billed outside standard subscriptions. The restored model kept its original pricing of $10 per million input tokens and $50 per million output tokens, a premium of roughly five times the introductory rate Anthropic set for its new Sonnet 5 model. Developers using Claude Code have also reported the system falling back to Opus 4.8 on some routine coding tasks, which is consistent with the tighter safety classifiers Anthropic described in its restoration announcement.
The episode is now being read across the industry as a preview of how governments will manage access to the most capable AI systems going forward. A frontier lab losing access to its own flagship product for more than two weeks, then negotiating a phased and metered return, sets a precedent that competing labs, including OpenAI and Google DeepMind, will need to account for in their own release planning. The bigger question hanging over the industry is whether this becomes a repeatable government review process or remains a one-time crisis response tied to this specific export dispute.
Source: Build Fast with AI | https://www.buildfastwithai.com/blogs/ai-news-today-june-29-2026
Claude Sonnet 5 Becomes the New Default Model, California Signs Statewide Anthropic Deal
Anthropic used the same evening that export controls lifted on Fable 5 to launch an entirely separate model. Claude Sonnet 5 debuted on June 30, 2026, and became the default model for every Free and Pro Claude user worldwide starting July 1. The model is also available to Max, Team, and Enterprise customers and through the API, priced at an introductory $2 per million input tokens and $10 per million output tokens through August 31, 2026, undercutting the standard rate for its predecessor, Sonnet 4.6.
What separates this release from a routine version bump is positioning. Anthropic is marketing Sonnet 5 as performing close to its flagship Opus 4.8 model on many tasks while running considerably cheaper, with improved coding, tool use, and reasoning capability plus lower measured rates of hallucination, sycophancy, and susceptibility to prompt injection compared with Sonnet 4.6. That combination of near-flagship capability at midsize pricing is squarely aimed at the agentic workflows enterprises are now trying to run at scale, from CRM updates to research and coding pipelines.
The same week, California became the largest state government AI deployment in Anthropic’s history, signing an agreement covering all state agencies at a 50 percent discount, structured as a volume arrangement for enrolled agencies rather than a flat per-user rate. Sonnet 5 is the primary model available under the agreement, with Opus 4.8 reserved for tasks requiring maximum accuracy, and the California Department of Employment Development is the first listed deployment partner, targeting workforce development and unemployment benefit processing. For an AI market watching how state governments choose vendors independent of federal policy, California just set the template that other states are likely to reference.
Source: unrot.co | https://unrot.co/blogs/today-top-10-ai-news-july-3-2026
OpenAI Delays GPT-5.6 Sol Rollout as Washington Demands Oversight
The most anticipated model release of the summer did not arrive on schedule. OpenAI is delaying the full public launch of GPT-5.6 Sol after the US government requested early access and additional oversight before broader availability, limiting initial access to a small group of vetted partners whose details have been shared with federal authorities. OpenAI has characterized the delay as temporary while it works with the administration on a repeatable release process, while also cautioning that government control over customer access should not become the industry standard.
Prediction markets had already priced in slippage before the government request became public. Tracking site FindSkill.ai noted that Polymarket odds for a GPT-5.6 release by June 28 collapsed from approximately 83 percent to about 18 percent as the June 22 to 27 window passed without an announcement, and the model instead entered a limited preview under the Sol, Terra, and Luna naming scheme reported by multiple outlets, including VentureBeat.
The timing lands the same week Anthropic’s own Mythos-class models were emerging from a separate government-triggered suspension, meaning both of the two most valuable AI startups in the country are now operating flagship releases under some degree of federal review. That is a meaningful shift from a year ago, when frontier model releases were purely commercial decisions. Enterprise buyers and marketing teams planning product launches around new model capabilities should expect release timing to become less predictable across the sector, not just at any single lab.
Source: MarketingProfs | https://www.marketingprofs.com/opinions/2026/55197/ai-update-july-3-2026-ai-news-and-views-from-the-past-week
Alphabet Sheds $269 Billion as DeepMind Talent Exodus Accelerates
Alphabet closed out the week of June 22 to 27, having lost roughly $269 billion in market value, driven largely by a cluster of senior departures from Google DeepMind to rival labs. Four high-profile researchers left within six days of each other: Noam Shazeer, a Transformer co-author and Gemini co-lead, moved to OpenAI on June 18; John Jumper, the Nobel laureate who led AlphaFold, moved to Anthropic on June 20; and Gemini AI coding lead Jonas Adler, along with Gemini pretraining specialist and AlphaFold contributor Alexander Pritzel, both moved to Anthropic on June 24.
The market reaction reflects a specific worry rather than generic AI jitters. Alphabet’s projected 2026 AI capital expenditure is approaching $190 billion, and investors are pricing the risk that spending at that scale on infrastructure means little if the researchers who translate compute into frontier capability keep leaving for competitors. Despite the selloff, the sell-side has not abandoned the stock: 28 of 33 analysts covering Alphabet maintained Buy ratings, pointing to a $460 billion contracted cloud backlog and 22 percent revenue growth as reasons the long-term case remains intact.
For the broader competitive landscape, this is a reminder that compute and capital are necessary but not sufficient conditions for frontier AI leadership. Anthropic, in particular, has now pulled top-tier scientific and engineering talent directly out of DeepMind’s ranks in the same month it was also navigating a government-mandated model shutdown, a combination that speaks to how contested the senior research talent market has become even amid regulatory turbulence.
Source: Build Fast with AI | https://www.buildfastwithai.com/blogs/ai-news-today-june-27-2026
Venture Capital Roundup: Together AI, TwelveLabs, Quantum Systems, and ElevenLabs Chase Fresh Capital
Global startup funding hit a record $510 billion in the first half of 2026, according to Crunchbase data reported by SiliconANGLE, with AI absorbing the overwhelming majority of new capital and OpenAI and Anthropic alone accounting for a large share of the total. Against that backdrop, several AI companies closed or pursued significant rounds during this specific window.
Together AI closed an $800 million Series C led by Aramco Ventures on July 1, and video AI company TwelveLabs raised a $100 million Series B led by NEA the same day. German autonomy and defense company Quantum Systems raised a $1.2 billion in Series D funding, giving it substantial new capital for AI-enabled defense hardware. Voice AI company ElevenLabs is exploring an employee stock sale at a reported $22 billion valuation, according to Reuters, citing Bloomberg reporting. Separately, 3D foundation model company Tripo AI raised another $150 million just a month after a prior $200 million raise, and privacy-focused AI company Venice.ai, led by crypto veteran Erik Voorhees and Seattle-based Jesse Proudman, raised $65 million, according to GeekWire.
| Company | Round | Amount | Date | Lead Investor |
|---|---|---|---|---|
| Together AI | Series C | $800 million | July 1, 2026 | Aramco Ventures |
| TwelveLabs | Series B | $100 million | July 1, 2026 | NEA |
| Quantum Systems | Series D | $1.2 billion | July 2 to 3, 2026 | Undisclosed |
| Tripo AI | Follow-on round | $150 million | Early July 2026 | Undisclosed |
| Venice.ai | Growth round | $65 million | Early July 2026 | Undisclosed |
| ElevenLabs | Employee stock sale | Valuation of $22 billion | Early July 2026 | N/A |
The pattern investors are pricing is capital concentration rather than broad distribution. Menlo Ventures raised a $3 billion fund built largely on the strength of its Anthropic stake, illustrating how frontier lab investment returns are now reshaping venture firm strategy itself. Later-stage AI application startups increasingly compete for a shrinking share of limited partner capital against the fundraising gravity of OpenAI and Anthropic, a dynamic that is likely to keep consolidating around infrastructure, defense, and vertical healthcare AI plays through the second half of the year.
Source: TechStartups | https://techstartups.com/2026/07/03/top-tech-news-today-july-3-2026/
CMS Launches ACCESS Model as Colorado’s AI Act Takes Effect for Healthcare Providers
Two significant healthcare AI compliance dates landed in the same window. The Centers for Medicare and Medicaid Services Innovation Center began enrolling its first cohort under the ACCESS Model, short for Advancing Chronic Care with Effective, Scalable Solutions, on July 1, 2026. The voluntary, ten-year payment model rewards Medicare Part B providers with recurring payments for using technology-enabled services, including AI tools, to manage chronic conditions and coordinate care, and it works alongside the FDA’s related TEMPO Pilot, which offers participating device manufacturers a form of regulatory relief from standard premarket authorization requirements while their devices operate within the model.
On the state side, Colorado’s amended AI Act took effect June 30, 2026, following more than a year of revision by the Colorado AI Policy Work Group convened by Governor Jared Polis. The updated law broadens an existing HIPAA carve-out so that most healthcare entities using AI for non-high-risk clinical recommendations only need to provide patients with a general notice that advanced technology is in use, rather than facing the full transparency and impact assessment obligations that apply to high-risk AI use elsewhere in the statute. Entities using AI to determine a patient’s eligibility for financial assistance still face additional disclosure requirements.
Together, these two developments capture the current split in US healthcare AI policy: federal agencies are building financial incentives and regulatory sandboxes to accelerate adoption, while states are independently setting their own disclosure and liability rules. For health systems and digital health vendors, that means compliance now requires tracking both a federal payment model timeline and a patchwork of state statutes that do not always align, a burden that is likely to intensify as more states finalize their own AI health rules later in 2026.
Source: Manatt, Phelps & Phillips | https://www.manatt.com/insights/newsletters/health-highlights/manatt-health-health-ai-policy-tracker
Google Loses Final EU Appeal Over 4.1 Billion Euro Android Antitrust Fine
Europe’s top court upheld a 4.1 billion euro antitrust fine against Google tied to its Android operating system, ending the company’s final appeal in a case that found Google used Android’s market power to reinforce the dominance of Search and Chrome and limit competition on mobile devices. The ruling closes out one of the longest-running Big Tech antitrust cases in European history.
Beyond the financial penalty, the decision matters because it reinforces the legal foundation European regulators are now using to pursue newer platform cases under the Digital Markets Act, several of which touch directly on AI features bundled into dominant operating systems and browsers. A court willing to uphold a multibillion euro fine against one of the world’s largest technology companies signals to regulators that similar enforcement actions against AI-driven bundling practices are legally durable rather than vulnerable to reversal on appeal.
For AI companies operating in Europe, the ruling is a reminder that the region’s regulatory posture toward dominant platforms has not softened even as AI competition intensifies globally. Companies building AI features into operating systems, browsers, or app stores with significant market share should expect European regulators to apply the same bundling and self-preferencing scrutiny that produced this outcome, particularly as the Digital Markets Act’s provisions increasingly intersect with generative AI product design.
Source: TechStartups, citing the Associated Press | https://techstartups.com/2026/07/03/top-tech-news-today-july-3-2026/
State Lawmakers Advance AI Chatbot Safety Bills in New York, New Jersey, and California
State legislatures used the final days before their July 4 recesses to push through a wave of AI-specific bills, with child safety emerging as the dominant theme. New York’s legislature approved a kids chatbot safety bill, S 9051, that prohibits AI chatbots from using features considered unsafe for minors and creates a private right of action, alongside a companion Artificial Intelligence Training Data Transparency Act requiring generative AI developers to publish high-level summaries of their training datasets. Both bills now await Governor Kathy Hochul’s signature, with a deadline of December 31, 2026.
New Jersey lawmakers gave two-chamber approval to the Kids Code Act, establishing new privacy, security, and parental control requirements for social media platforms that increasingly rely on AI-driven recommendation and chat features. California Governor Gavin Newsom signed AB 2148 on June 30, 2026, a narrower but symbolically important bill clarifying that public school employees and contractors must be actual human beings, a direct response to schools experimenting with AI-driven instructional roles. Not every state moved in the same direction: Arizona Governor Katie Hobbs vetoed three Republican-passed AI bills in late June as part of a mass veto of 88 bills in a single day, with no public explanation of her specific objections.
The throughline across these state actions is that child safety, not general-purpose AI regulation, is now the fastest-moving legislative category nationwide, reflecting bipartisan pressure that transcends the usual state-by-state partisan divide on tech regulation. Companies building consumer-facing AI chatbots, especially those with any plausible reach to minors, should treat 2026 state legislative sessions as an active compliance risk rather than a background policy trend, since several of these bills carry private rights of action that create direct litigation exposure independent of any regulatory enforcement action.
Source: Transparency Coalition | https://www.transparencycoalition.ai/news/ai-legislative-update-july3-2026
CISA Flags Critical LiteLLM Vulnerability Exposing Enterprise AI Gateways
The Cybersecurity and Infrastructure Security Agency added CVE-2026-42271 to its Known Exploited Vulnerabilities catalog on June 27, 2026, flagging an unauthenticated remote code execution chain in LiteLLM’s AI Gateway. LiteLLM is a widely deployed open source proxy that lets developers route requests to multiple large language model providers through a single unified API, making it a common infrastructure inside enterprise AI stacks. The vulnerability exploits Model Context Protocol endpoints to gain full access to the server environment, including any OpenAI and Anthropic API keys configured on that gateway.
The severity here compounds rather than isolates. A successful exploit does not just hand an attacker code execution on the LiteLLM server itself; it harvests credentials for every AI provider the gateway was configured to reach, turning a single infrastructure flaw into a multi-vendor credential breach. A patch was released on June 20, 2026, ahead of the CISA advisory, and organizations running LiteLLM in production are being urged to update immediately, rotate every API key configured in the deployment, and audit access logs for suspicious activity predating the patch.
The disclosure lands amid a broader spike in AI-related vulnerability research. Epoch AI reported that AI-assisted vulnerability discovery drove a record spike in June 2026, with 21 organizations disclosing roughly 1,500 high and critical severity CVEs during the month. For chief information security officers, the LiteLLM incident is a concrete illustration of why AI gateway infrastructure, not just the models themselves, now needs to sit inside the standard enterprise vulnerability management program rather than being treated as a lightweight developer convenience layer.
Source: Build Fast with AI | https://www.buildfastwithai.com/blogs/ai-news-today-july-1-2026
Salesforce Data Shows Enterprise Agentic AI Adoption Nearly Doubling in a Year
Enterprise deployment of agentic AI moved decisively from pilot projects to full production this week. A Salesforce survey found that service organizations’ agentic AI adoption rose from 39 percent to 66 percent over the past year, with 70 percent of adopters reporting measurable returns within 60 days of deployment. The shift is visible in specific enterprise rollouts: Warner Bros. Discovery is rebuilding its advertising technology stack on AWS agents, Yahoo launched an open agent network spanning 23 partners, and WPP is testing AI buyer agents for video ad placement.
The adoption numbers come with an important caveat from separate research cited in the same reporting cycle: only about 7 percent of companies currently run fully autonomous AI agents in production, meaning the surge in reported adoption still mostly describes supervised or partially autonomous deployments rather than end-to-end automation. Practitioner guidance circulating this week also warned that many agentic AI projects remain at risk of cancellation because teams are bolting agents onto existing processes instead of redesigning workflows around them, and are failing to set clear baselines for time savings, error rates, and human effort before declaring success.
For go-to-market and operations leaders, the signal is that agentic AI has crossed from side experiment to credible infrastructure, and competitors that have not yet automated planning, trafficking, or optimization loops risk falling behind on cycle time and headcount efficiency. At the same time, the gap between adoption headlines and genuinely autonomous production use suggests the real competitive advantage over the next year will go to teams that treat agentic AI as a business architecture problem rather than a simple model integration.
Source: AI Agent Store | https://aiagentstore.ai/ai-agent-news/daily/2026-06-28
Grok 4.5 Enters Private Beta as Chinese Open Source Model Closes the Gap
Elon Musk announced on X on June 28, 2026, that Grok 4.5 has entered private beta at SpaceX and Tesla, an internal testing arrangement that keeps the model out of public hands while xAI validates it inside two of Musk’s other companies first. Reporting on the model describes a roughly 1.5 trillion parameter architecture, positioning it as a significant scale increase over prior Grok generations.
At the same time, competitive pressure from Chinese labs intensified. Z.ai’s GLM-5.2, an open-weight model, has climbed developer rankings after demonstrating coding and agentic capabilities that approach Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5 at a fraction of the licensing cost, attracting startups looking to cut AI infrastructure spending and reduce dependence on proprietary US models. A separate 1.6 trillion parameter Chinese model trained entirely on domestic chips was also open-sourced under an MIT license during this window, underscoring how quickly China’s frontier open source ecosystem is narrowing the capability gap with closed Western labs even as US export controls on advanced chips remain in place.
Enterprise adoption of these open-weight Chinese models in the US and Europe continues to face real friction, including data security concerns and regulatory hesitation about incorporating Chinese-developed models into core infrastructure. But the trend line is unmistakable: lower cost, frontier-adjacent open weight models are compressing the pricing power that closed US labs have enjoyed, and marketing and product teams evaluating AI vendors should expect that pricing pressure to intensify vendor competition through the rest of 2026.
Source: MarketingProfs | https://www.marketingprofs.com/opinions/2026/55197/ai-update-july-3-2026-ai-news-and-views-from-the-past-week
Palantir’s Alex Karp Blasts Frontier AI Pricing as Nvidia Nemotron Gains Ground
Palantir CEO Alex Karp used a CNBC interview on July 2, 2026, to sharply criticize how frontier AI labs price their products, framing the gap between proprietary model pricing and lower-cost alternatives as effectively a wealth transfer to the companies selling AI tools rather than the businesses using them. Karp specifically contrasted the pricing of frontier Western models, including GPT-5.5 at roughly $15 per million output tokens and Claude Sonnet 5 at its introductory $10 rate, against Nvidia’s Nemotron 3 Ultra model, which Palantir is positioning as a lower-cost alternative for government contracts.
The interview doubled as a positioning exercise for Palantir’s own AI Platform, which the company is pairing with Nvidia’s Nemotron integration to offer government and enterprise customers a path around dependency on frontier labs for every workload. Karp’s comments landed the same week Nvidia rolled out a revenue-sharing infrastructure model that several industry newsletters flagged as one of the more consequential AI business model shifts of the summer, alongside SoftBank establishing a new US neocloud unit called SB Neo.
The commentary reflects a widening rift in how the AI supply chain talks about value capture. Frontier labs are betting that capability differentiation justifies premium pricing even as competition intensifies, while infrastructure and defense-adjacent players like Palantir are betting that most enterprise and government workloads do not require frontier-level intelligence at frontier-level prices. How that argument resolves will shape which vendors win the largest government AI contracts over the next several budget cycles, particularly as agencies weigh cost against capability for non-critical workloads.
Source: Build Fast with AI | https://www.buildfastwithai.com/blogs/ai-news-today-july-4-2026
What This Week Means for the AI Industry
The nine days covered here confirm that AI’s center of gravity has shifted from pure model capability to who controls access, capital, and compliance around that capability. Export controls took two flagship models offline and back again, a record half-year of venture funding kept concentrating around a handful of infrastructure and frontier players, and healthcare and state regulators moved on separate but overlapping timelines that will define compliance work for the rest of 2026.
Watch the July 7 expiration of Fable 5’s subscription-included access, Colorado’s early enforcement actions under its amended AI Act, and whether OpenAI’s delayed GPT-5.6 rollout produces a repeatable government review model or another one-off standoff.
Frequently Asked Questions
What is the biggest AI news story between June 27 and July 4, 2026?
The most consequential story is Anthropic’s restoration of Claude Fable 5 and Mythos 5 on July 1, 2026, after an 18-day shutdown under a US Department of Commerce export control order, closely followed by the same-week launch of Claude Sonnet 5 as Anthropic’s new default model.
Why was Claude Fable 5 taken offline, and is it back now?
Fable 5 and Mythos 5 were suspended on June 12, 2026, under an emergency federal export control directive. The Commerce Department lifted the controls, and Anthropic restored global access starting July 1, though Fable 5’s inclusion within standard subscription limits is only guaranteed through July 7, 2026.
What is Claude Sonnet 5, and how much does it cost?
Claude Sonnet 5 is Anthropic’s new midsize agentic model, launched June 30, 2026, and made the default for all Free and Pro Claude users on July 1. It is priced at an introductory $2 per million input tokens and $10 per million output tokens through August 31, 2026.
Why did Alphabet’s stock drop in late June 2026?
Alphabet lost roughly $269 billion in market value after four senior Google DeepMind researchers, including Transformer co-author Noam Shazeer and AlphaFold lead John Jumper, departed for OpenAI and Anthropic within six days, raising investor concern about Alphabet’s ability to convert its nearly $190 billion 2026 AI capital budget into frontier capability.
Is GPT-5.6 available to the public yet?
Not fully. OpenAI delayed the public rollout of GPT-5.6 Sol after a request from the US government for early access and additional oversight, limiting initial availability to a small group of vetted partners as of early July 2026.
What is the CMS ACCESS Model, and how does it affect healthcare AI?
The ACCESS Model is a voluntary, ten-year Medicare payment program from the CMS Innovation Center that began enrolling its first cohort on July 1, 2026. It rewards Medicare Part B providers for using technology-enabled, including AI-driven, services to manage chronic conditions and coordinate patient care.
Why did the EU fine Google over Android, and did the fine survive an appeal?
The European Union previously fined Google 4.1 billion euros for using Android’s market position to reinforce Search and Chrome and limit competitor access to mobile devices. Europe’s top court upheld that fine in early July 2026, closing out Google’s final appeal.
What new AI laws take effect for healthcare providers in 2026?
Colorado’s amended AI Act took effect June 30, 2026, adjusting HIPAA-related disclosure obligations for healthcare entities using AI in non-high-risk clinical recommendations. Several other states, including New York and New Jersey, are separately advancing AI chatbot safety bills that could affect healthcare-adjacent consumer products.
What is GLM-5.2, and how does it compare to Claude and GPT models?
GLM-5.2 is an open-weight model from Chinese AI company Z.ai that has demonstrated coding and agentic capabilities approaching Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5, at a significantly lower cost, though enterprise adoption in the US and Europe remains limited by data security and regulatory concerns.
Is Grok 4.5 publicly available?
No. As of early July 2026, Grok 4.5 is in private beta, limited to internal testing at SpaceX and Tesla, and xAI has not announced a public release date.
