Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect

techcrunch.com Oct 9, 2026

Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect

Curated by Shubham · TechCapsules · Oct 9, 2026

AI-summarised brief · reviewed before publication

OpenAI dismissed three safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—after an investigation found they allegedly shared confidential data with an external AI safety group. The researchers released an open letter denying the misconduct claims, arguing their firings signal a chilling effect on the company’s culture of open safety dialogue. They contend that their collaboration with outside experts was essential for addressing AI risks and that the abrupt terminations undermine transparency and accountability within OpenAI.

💡 Why It Matters

· The firings raise concerns about how AI firms balance proprietary secrecy with external safety oversight, potentially stifling critical independent scrutiny of emerging technologies.

AI Capsules OpenAI

Read full article on techcrunch.com

Smart Contact Lenses May Arrive as Soon as 2030. Here’s What They…

Dubai-based Xpanceo has developed AI‑powered smart contact lenses that integrate a microdisplay, biosensors, and a microbattery, paired with a head‑mounted companion device for power, computing, and wireless connectivity. The lenses can display information, verify identity, provide augmented reality overlays, and monitor health biomarkers such as glucose, proteins, electrolytes, stress hormones, and intraocular pressure via tear fluid analysis. With 28 prototypes and 50 patents, Xpanceo plans to unveil the system at Mobile World Congress 2027, targeting a 2030 consumer release. The technology promises continuous, non‑invasive health monitoring and expanded data access directly on the eye.

💡 By enabling real‑time biochemical sensing without blood draws, the lenses could transform diabetes management and early eye‑disease detection, offering a seamless, proactive health solution that current wrist‑based wearables cannot match.

techcrunch.com · Oct 9

Popular AI leaderboard Arena nearly doubles valuation to $3.1B valuation in 10…

Arena, a UC Berkeley‑originated AI model ranking platform, secured a $200 million Series B, valuing the company at $3.1 billion—nearly double its $1.7 billion valuation a decade earlier. The funding, led by Lightspeed and Khosla, follows a $150 million Series A and a $100 million annualized revenue milestone. Arena offers a free crowdsourced leaderboard and a commercial AI Evaluations service, adding an alignment category to assess safety and honesty in model outputs. The round underscores growing demand for neutral, real‑world AI performance metrics.

💡 The surge in Arena’s valuation signals a shift toward community‑driven AI evaluation, offering enterprises a reliable benchmark amid model gaming. This move could redefine how safety and alignment are measured in the rapidly evolving AI landscape.

taipeitimes.com · Oct 9

Do not pave the way for competitors

Elon Musk has engaged Taiwan Semiconductor Manufacturing Co. (TSMC) to discuss a partnership for his Terafab project, a planned Texas facility aimed at producing over 1 terawatt of AI computing power annually. TSMC is evaluating the investment, but the partnership’s structure, ownership, manufacturing processes, and timeline remain undecided. Terafab would integrate logic, memory, advanced packaging, testing, and photomask production under one system, requiring 2‑nanometer Gate‑All‑Around technology. The arrangement could shift TSMC from a customer to a potential competitor, redefining the foundry’s competitive moat.

💡 The deal could give Musk’s companies unprecedented control over cutting‑edge chip production, altering supply‑chain dynamics and potentially eroding TSMC’s long‑standing manufacturing advantage.

namibian.com.na · Oct 9

AI makes scammers harder to spot

Cybersecurity specialists warned at the Cybersafety for All seminar in Windhoek that personal data posted on social media is being harvested to craft hyper‑personalised phishing attacks. Andrew Brown highlighted AI‑driven real‑time face‑swapping, enabling criminals to impersonate individuals during video calls and to bypass “view‑once” privacy controls by photographing screens. Sheldon Shava explained that platforms such as Facebook and Instagram supply the details needed to build victim profiles, which are then weaponised in targeted scams. Graduate Tafadzwa Mhuriyengwe added that AI can generate polished, convincing emails and messages, making fraudulent communications harder to spot. Speakers urged users to limit online disclosures, verify unusual requests, and urged organisations to regularly test and harden their security systems.

💡 AI‑enhanced impersonation erodes trust in digital interactions, forcing both individuals and businesses to adopt stricter verification practices.

prnewswire.com · Oct 9

$2.6 Trillion and Climbing: Where the AI Spending Is Going

Worldwide artificial‑intelligence spending is projected to hit $2.59 trillion in 2026, a 47 percent jump from 2025, and to rise to $3.49 trillion in 2027, according to Gartner. AI infrastructure accounts for $1.43 trillion of the 2026 total, representing more than 45 percent of all AI outlays. Gartner’s John‑David Lovelock calls 2026 the “inflection year.” The surge is driven by large cloud and internet platforms expanding data‑center capacity and chip makers supplying processors, custom accelerators and networking gear. Companies such as Microsoft, Meta, Broadcom and AMD are reporting double‑digit revenue growth and capital‑expenditure commitments exceeding $100 billion annually. The Nasdaq Composite is up roughly 17 percent so far in 2026 as investors reward firms that build and sell AI hardware.

💡 The concentration of spending on a narrow supply chain creates a high‑stakes battleground where a few hardware vendors can capture disproportionate returns.

techradar.com · Oct 9

Could .chatgpt and .claude be the next generation of web domains? ICANN…

The Internet Corporation for Assigned Names and Numbers (ICANN) has disclosed details of the 1,615 applications it received in its latest round for new generic top‑level domains (gTLDs). Of those, 333 are brand‑related, with major tech firms seeking to secure domains that reflect their products and services. More than half of the submissions—over 800—originated from North America, while Europe contributed 506, together accounting for roughly 85 % of all applications. The filing pool also includes 21 internationalized domain name (IDN) requests in non‑Latin scripts such as Arabic and Chinese. Notable applicants include OpenAI, which filed 15 proposals including .chatgpt; Google, with over 30 requests such as .api and .agent; Meta, seeking .superintelligence and .wearables; and Anthropic, applying for .anthropic and .claude. The competition is especially fierce for AI‑related strings. ICANN will hold a Prioritization Draw in early 2027, with any approvals expected to be announced in the latter half of that year.

💡 AI‑centric domains could become the new branding battleground, giving tech giants direct control over how users locate and trust their services online.