AI News July 14, 2026: SK Hynix Plunges After $26.5B Nasdaq Debut, xAI Halts Grok Build Uploads, Open-Source Projects Reject AI Code
Today in AI: SK Hynix’s historic $26.5 billion Nasdaq debut ended in a 9% plunge on day two as AI chip sentiment reversed hard. xAI shut down Grok Build’s default code repository uploads after a 5.1 GB data leak. A prominent AI researcher predicted open-weight models above GPT-5.5 tier will be banned within six months. And the open-source community is in full revolt against AI-generated code contributions.
SK Hynix Plunges 9% After Record $26.5 Billion Nasdaq Debut
SK Hynix, the South Korean chipmaker behind the HBM memory inside nearly every advanced AI accelerator, made the largest ADR offering in market history at $26.5 billion when it listed on Nasdaq on July 10 under the ticker SKHY — bigger than Alibaba’s 2014 debut. Institutional anchors committed roughly $7 billion before the first trade, and the stock surged on opening day.
But by midday July 13, SKHY fell around 9% to $139.14, giving back most of its opening gains. TheStreet reports the pullback reflects the tension between SK Hynix’s 58% share of the global HBM market and the reality that the stock arrived after a 710% twelve-month rally in its Korean shares plus a recent 25% drawdown. GraniteShares launched leveraged ETFs (SKUU and SKDD) on July 13, the first trading day after the debut.
xAI Halts Grok Build Code Repository Uploads After 5.1 GB Data Leak
xAI disabled Grok Build’s default behavior of uploading code repositories to its servers following the discovery of a 5.1 GB data leak, according to HuggingNews. The feature, which uploaded user code by default without explicit opt-in, exposed project files from users who hadn’t changed their settings. xAI applied a server flag to halt the behavior and is investigating the scope of the exposure.
The incident highlights a recurring tension in AI coding tools: the trade-off between providing better context (more code uploaded = better suggestions) and user trust (developers don’t expect their private repos to be uploaded without clear consent). Cursor, Copilot, and Claude Code have all navigated this boundary differently, with varying levels of transparency about what data leaves the developer’s machine.
AI Researcher Predicts Open-Weight Model Ban Within Six Months
Nathan Lambert, a prominent AI researcher, published a prediction that the US will move to ban or restrict open-weight AI models above the GPT-5.5 / Claude Opus 4.8 / GLM-5.2 capability tier within roughly six months. He cited sources pointing at White House executive-order discussions. The same weekend, Zhipu founder Tang Jie argued in a public memo that frontier AI must remain open, and shipped GLM-5.2 free to download and commercialize.
The debate reflects a fundamental split: national security advocates argue open-weight frontier models are too dangerous to distribute freely, while open-source advocates argue that restricting them hands a monopoly to a handful of closed-source companies. The outcome will reshape who can build with the most capable AI and how — directly affecting every developer and startup in the AI ecosystem.
Open-Source Projects Revolt Against AI-Generated Code Contributions
A growing number of major open-source projects are actively rejecting AI-generated code. cURL creator Daniel Stenberg shut down the project’s six-year bug bounty program in January after $86,000 in payouts, citing that valid reports had dropped to roughly 5% as AI-generated submissions flooded in. Ghostty now explicitly bans unapproved AI code, and tldraw auto-closes every external pull request.
The FT framed the trend as a “tragedy of the commons” — vibe coders ship faster while unpaid maintainers inherit the triage burden. As AI makes it trivially easy to generate code and bug reports at scale, the cost of sorting signal from noise falls disproportionately on maintainers who are already stretched thin. This tension will only intensify as AI coding tools become more capable.
Trump Administration Discusses Framework to Fast-Track Open-Source AI Releases
In a seemingly contradictory move to Lambert’s prediction, HuggingNews reports that the Trump administration and AI industry leaders are discussing a framework to fast-track open-source AI model releases. The framework would reportedly create a streamlined approval process tied to specific safety benchmarks, allowing models that pass defined thresholds to be released openly without extended government review periods.
The dual signals — Lambert warning of an imminent ban while the administration discusses fast-tracking — illustrate how fractured US AI policy has become. Different factions within the same government are pulling in opposite directions, creating uncertainty for every AI company deciding whether to open-source their next model.