OpenAI Dismisses Three Safety Researchers After Confidentiality Investigation | AI News

Nic Reeve14:00 02.10.20265 min read

OpenAI said on Thursday, October 1, 2026, that it had fired three researchers after an internal investigation found they mishandled sensitive company information, including material allegedly shared with an outside AI-safety organization. The dismissals, first reported by The Wall Street Journal and later confirmed by OpenAI to several outlets, are now a major AI news story because at least two of the employees worked on safety and alignment research.

What did OpenAI announce on October 1?

OpenAI confirmed that three employees had left the company following an investigation into how they accessed and handled confidential information. The company did not initially identify the employees or disclose the documents involved. OpenAI described the case as a policy violation, not as a dispute over the substance of safety research.

According to OpenAI’s statement reported by the BBC on October 2, 2026, the company fired three researchers.
According to OpenAI, the employees violated rules governing access to and handling of sensitive information.
According to The Wall Street Journal, the alleged sharing involved a third-party organization working on AI safety.

An OpenAI spokesperson told The Wall Street Journal, “We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information.” The spokesperson added, “Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.”

Who were the researchers?

Early reports did not name the dismissed employees. A later report carried by Agence France-Presse identified them as Jasmine Wang, Tomek Korbak and Mikita Balesni, citing reporting from The Wall Street Journal. OpenAI has not publicly released a detailed account of each person’s role or the specific conduct attributed to each researcher.

According to Agence France-Presse on October 2, 2026, the three names reported were Jasmine Wang, Tomek Korbak and Mikita Balesni.
According to the BBC, at least two of the three employees had worked in safety research.
According to South Korea’s Seoul Economic Daily, one employee handled support and communications involving Redwood Research and METR.

The reporting places the case inside OpenAI’s safety and alignment work. Alignment research seeks to make AI systems follow intended goals and avoid harmful behavior. That connection gives the dismissals wider importance than a routine personnel action, although the available reports do not establish that the researchers were fired for raising safety concerns.

What information was allegedly shared?

OpenAI has not said whether the material consisted of model evaluations, research findings, internal messages, security details or another category of company data. News reports describe the information only as confidential or sensitive. The identity of the outside organization has also not been confirmed by OpenAI.

According to Tech Times on October 2, 2026, OpenAI had not specified whether the material involved model capabilities, testing results, internal communications or research.
According to The Wall Street Journal, the recipients were connected to a third-party AI-safety organization.
According to Seoul Economic Daily, reporting linked one researcher’s work with Redwood Research and METR.

Redwood Research studies advanced AI systems and safety questions. METR, formerly known as the Machine Intelligence Research Institute’s evaluation-related group in some public discussions, is known for work assessing the capabilities and risks of frontier models. The available reporting does not say that either organization received the allegedly mishandled material.

Why does the case matter to AI safety teams?

The dismissals expose a tension inside frontier AI companies. Safety researchers often work with outside nonprofits, academics and evaluators because independent testing can reveal weaknesses that internal teams miss. Companies also restrict access to confidential data because model information, security procedures and unreleased research can create commercial and safety risks if distributed without authorization.

OpenAI’s public explanation focused on process. The company said the employees went outside established procedures and breached internal trust. The reports do not show that OpenAI has accused the named researchers of selling information, helping a commercial rival or causing a public data breach.

The distinction matters. Sharing information with an external safety group may involve legitimate research collaboration, unauthorized disclosure, or both, depending on what was shared and whether approval existed. The facts available on October 2 do not resolve that question.

What remains unknown about the investigation?

OpenAI has confirmed the number of dismissals and the broad policy rationale, but it has not published the investigation’s findings. The company has also not said when the alleged disclosures occurred, how the information left OpenAI, which policies applied, or whether the external organization received the material directly.

According to OpenAI’s statements reported by the BBC and CBS News, the company has not publicly identified the affected documents.
According to CBS News on October 1, 2026, The Wall Street Journal was the first outlet to report the departures.
According to Agence France-Presse on October 2, OpenAI did not confirm the researchers’ identities in its own statement.

No public report cited in the current coverage says that prosecutors opened a criminal investigation or that regulators have announced action. The reports also do not indicate whether the three researchers plan to challenge their dismissals.

What happens next for OpenAI and its safety work?

OpenAI will face pressure to explain how independent safety collaboration can continue while confidential information remains protected. Researchers, employees and outside evaluators will likely seek clearer rules on approved disclosures, documentation and escalation channels. The company has not announced new procedures or a review of its broader safety program.

The immediate timeline is short:

According to The Wall Street Journal, OpenAI notified employees about the departures before the report published on October 1, 2026.
According to TechCrunch on October 1, OpenAI confirmed that it had parted ways with three researchers.
According to the BBC on October 2, OpenAI continued to characterize the conduct as mishandling sensitive information outside approved procedures.

Further details will depend on statements from the three researchers, the outside safety organization and OpenAI. Until those accounts emerge, the confirmed record is limited to three dismissals, an internal investigation and allegations involving unauthorized handling of sensitive information.

Sources

AI & Tech

AInews Weekly: Education Gaps, Datacenter Surge and Rutgers Trust Study

AInews Weekly: Education Gaps, Datacenter Spending Spike, and New Public Trust Data During the week of September 11, 2026, AInews stories ranged from a sweeping DataCamp survey on classroom AI use to fresh IDC numbers on infrastructure spending and new Rutgers research on public trust in automated decision systems, showing how fast artificial intelligence is spreading while core skills and guardrails struggle to keep pace. What did DataCamp reveal about AI in classrooms in 2026? DataCamp’s new “AI in Education” report, released on September 10, 2026, found that student use of AI tools is now near universal, while fluency and critical-thinking safeguards lag behind. The study surveyed more than 150 teachers and 150 students, highlighting a sharp divide between everyday AI use and formal guidance. According to DataCamp’s 2026 AI in Education report, published via Business Wire and covered by the Las Vegas Sun, key findings include: Scope: More than 150 teachers and more than 150 students across different schools were surveyed about AI use and attitudes. Adoption: The report describes student AI adoption as effectively universal among respondents, meaning most students rely on AI tools in some form for schoolwork. Skills gap: DataCamp concludes that “massive gaps remain between AI adoption and fluency,” with many students using tools they do not fully understand. Critical thinking worries: Educators in the survey express concern that over‑reliance on AI may weaken students’ independent reasoning and writing skills. Policy uncertainty: Respondents report uneven or unclear school protocols for AI use, from plagiarism rules to allowed tools during assignments. DataCamp positions itself as an AI and data upskilling platform and says the report is meant to give educators a baseline for how the “first AI‑native class,” graduating in 2026, is actually using automation in its daily work. The company argues that structured training in topics such as AI ethics, data literacy, and prompt design is now a prerequisite for meaningful classroom use rather than an optional add‑on. Earlier in 2026, DataCamp pledged free AI training for one million teachers and students worldwide through its DataCamp Classrooms program, including courses in Python, SQL, Power BI and broader AI literacy. The September education report puts numbers and concern behind that pledge, framing it as a response to the specific gaps the survey identified. How is DataCamp expanding AI tools and content for professionals and organizations? DataCamp spent Q3 2026 pushing AI deeper into its corporate and professional learning products, from an expanded AI Tutor interface to AI Adoption Insights dashboards for team admins. The platform also rolled out new tracks tied to OpenAI models, Anthropic’s Claude, and LangChain‑based AI engineering training. In its Q3 2026 roadmap webinar, summarized on DataCamp’s site, the company reported major content and feature milestones across the first half of the year: New content: More than 120 new courses, 21 new learning tracks, and support for 13 languages added in the first half of 2026. AI Tutor expansion: DataCamp renamed its “AI Native” learning mode to **AI Tutor** and began integrating Anthropic’s Claude and Claude Cowork directly into that experience. Infrastructure: A DataCamp MCP server connects Claude to the platform, letting admins manage learning plans and pull reports through natural‑language prompts. Analytics: “AI Adoption Insights” in Group Hub show how teams use AI tools day to day and benchmark that usage against other organizations. Specialized tracks: New tracks focus on Claude fundamentals, Claude for software engineers, and token cost management for developers, along with courses for Microsoft Fabric, Power Platform, Polars, and Apache Airflow. Certifications: A Python Developer Associate certification is live, with AI for Business, AI Agent Fundamentals, and AI Leadership credentials scheduled to round out the AI fluency lineup. Earlier in the year, DataCamp also announced a partnership with LangChain to launch an “AI Engineering with LangChain” track, aimed at software developers who want to build production‑grade AI applications. That track is positioned as part of the broader move from basic prompt skills to full AI engineering, covering topics such as chaining tools, handling context windows, and monitoring model behavior. The new courses build on DataCamp’s coverage of frontier models, including blog analysis of OpenAI’s GPT‑6 “Astra” launch and comparison pieces that try to map when developers should choose newer OpenAI systems over competitors like Anthropic’s Claude Fable 5.1. Together with AI Tutor and LangChain tracks, these updates show DataCamp targeting both the education market and working engineers with more intensive AI workflows. What does IDC report about AI‑driven infrastructure and networking spending? IDC’s latest infrastructure research points to sharp growth in networking hardware as organizations build out AI data centers. The firm highlights a 43.4% year‑over‑year surge in the Ethernet switch market to $18.9 billion in the second quarter of 2026, driven largely by AI training and inference workloads. According to IDC’s August 2026 networking market blog post: Ethernet switch revenue rose 43.4% year over year in Q2 2026 to reach $18.9 billion, which IDC links directly to demand from AI datacenters. Most of this growth comes from high‑end switches deployed in hyperscale and large enterprise facilities running GPU‑dense AI clusters. IDC analysts argue that AI workloads are changing network design, pushing vendors toward higher port densities and new designs optimized for large‑scale parallel processing. The report suggests that spending on AI infrastructure is no longer experimental and is instead driving record‑level datacenter budgets across sectors. Alongside networking, IDC’s resource center has highlighted moves such as NVIDIA’s acquisition of Hugging Face as part of a broader trend toward enterprise adoption of open models, with vendors racing to package open‑source and proprietary AI systems into consumable platforms. These combined trends show the business side of AI evolving beyond model releases into large hardware purchases, mergers and acquisitions, and long‑term infrastructure planning. How is Rutgers working with AI in libraries, research, and public sentiment? Rutgers University spent early September 2026 pushing both practical AI guidance and new research on public attitudes. The institution launched workshops through Rutgers Libraries on navigating AI tools and supported a Tech Xplore‑reported survey showing discomfort when AI makes decisions about people rather than simply assisting them. Rutgers’ official IT site describes a growing university‑wide AI initiative, spanning healthcare, data science, and library services. Within that framework, Rutgers Libraries announced “Navigating AI” workshops on September 2, 2026, with goals that include: Teaching students and staff how to evaluate AI tools and outputs for reliability and bias. Explaining how generative models handle data, privacy, and attribution. Showing library users how to blend AI search or summarization tools with traditional academic research methods. On September 9, 2026, Tech Xplore reported new Rutgers‑linked survey research into American attitudes toward artificial intelligence. The article states that: Survey participants are broadly comfortable with AI when it works as a tool they control, such as autocorrect or recommendation engines. Comfort levels drop sharply when AI shifts from assistance to decision‑making about people, for example in credit scoring, hiring, or predictive policing. Respondents express concern about transparency and fairness when AI systems make high‑stakes choices, even if they value the efficiency gains. Rutgers’ Wireless Information Network Laboratory (WINLAB) is also using September to host back‑to‑back technical workshops focused on advanced networking testbeds, including COSMOS3 on September 17, 2026, where AI‑driven optimization and traffic management form part of the agenda. While infrastructure workshops may seem distant from public‑trust surveys and library training, they illustrate how AI work at Rutgers spans basic research, user education, and social impact. What other notable AI developments rounded out this week’s landscape? The broader AI week around September 11 included frontier‑model debates, open‑model consolidation, and new tools for monitoring how organizations actually use AI. These events connect the education stories from DataCamp, the infrastructure spike identified by IDC, and the trust questions raised by Rutgers. Across various sources, notable developments included: Model launches: Commentators summarized the early September release of OpenAI’s GPT‑6 “Astra,” noting how the model’s arrival intensified discussion around data sovereignty and control in AI infrastructure. Marketplace consolidation: IDC’s coverage of NVIDIA’s move to acquire Hugging Face underscores how hardware vendors are seeking stronger positions in open‑model ecosystems used by enterprises. Usage analytics: DataCamp’s AI Adoption Insights aim to show organizations where AI is truly embedded in daily workflows, not just in pilot projects. Ethical guidance: Rutgers’ combination of public‑sentiment research and practical workshops highlight a growing institutional push to give ordinary users tools to judge when AI is being used appropriately. The original weekly round‑up published by Solutions Review on September 12, 2026, framed these updates as a snapshot of how fast AI is moving into mainstream systems while educators, IT teams, and researchers scramble to manage its consequences. By pulling together survey data, infrastructure spending figures, and new teaching programs, this week’s news shows the spread of AI across technical, social, and institutional lines—and how far formal governance still has to go.

Nic Reeve·14:00 12.09.2026

AI & Tech

Intel’s AI Surge Sparks Fierce Valuation Clash as Some Models Flag 80%+ Upside

Intel’s rapid push into artificial intelligence chips and foundry services has ignited a sharp debate over what the stock is really worth, with one widely followed narrative now implying a fair value near $500 per share — more than four times the recent market price. While mainstream analysts still cluster between roughly $90 and $120 per share, user-driven valuation models and some sales-based frameworks argue that the market is deeply underestimating Intel’s long‑term AI earnings power, leaving the stock potentially more than 80% below fair value. Where the 82% Undervaluation Claim Comes From The headline figure that Intel could be about 82% below fair value stems from a narrative used on retail‑focused valuation platforms, which apply aggressive growth and margin assumptions to Intel’s emerging AI businesses. In several recent notes, that framework points to a fair value around $500.93 per share , compared with a share price near the $90–$100 range in late August 2026. On that basis, Intel is framed as roughly 80–82% undervalued, with the gap driven by bullish expectations for x86 server CPUs, AI accelerators and foundry contracts over the next decade. These narratives typically assume: Strong, sustained growth in Intel’s Data Center and AI (DCAI) segment. High adoption of Intel’s advanced manufacturing nodes, such as 18A, by external foundry customers. AI‑linked revenue eventually commanding premium valuation multiples similar to leading GPU and cloud infrastructure providers. Critically, this $500+ fair value is not a consensus Wall Street target but a specific, scenario‑driven model that extrapolates current AI momentum far into the future. Intel’s Latest AI and Earnings Momentum The bullish valuation arguments have gained traction as Intel’s reported numbers show AI demand increasingly driving the business. For the second quarter of 2026, Intel reported revenue of about $16.1 billion, up 25% year over year , and adjusted earnings per share of $0.42, beating analyst expectations. The company’s Data Center and AI Group stood out, delivering approximately 59% year‑over‑year growth , with management noting that AI‑linked businesses grew more than 70% and now account for roughly 70% of total revenue. Intel also guided third‑quarter revenue to a range of $15.8 billion to $16.8 billion and gross margins in the low‑40% band, signaling confidence that AI‑related demand will remain robust despite broader concerns about chip valuations. On the strategic side, Intel highlighted signed foundry and advanced packaging agreements with major technology players, including Google, Nvidia, Tesla and Apple, alongside partnerships tied to its 18A manufacturing node and High NA EUV lithography. Foundry revenue rose by more than 30% year over year, although external customers still represent a small share of the segment, keeping the long‑term foundry thesis partly unproven. Mainstream Fair Value Estimates: 90–120 Dollar Range Traditional analyst research paints a far more moderate picture of Intel’s intrinsic value. Morningstar, which has repeatedly updated its Intel model in response to the AI boom, lifted its fair value estimate multiple times in 2026. Earlier in the year, analysts raised Intel’s fair value to $90 per share from $60, citing a “stunning” rise in server CPU demand and a growing AI infrastructure build‑out. Following stronger results and upgraded expectations, Morningstar later increased its fair value estimate to around $105 per share , and some commentary mentions fair value figures just above $100 as AI‑related assumptions were refined further. Other analyst summaries show valuation targets and fair value estimates clustering between roughly $88 and $115 per share , with some firms setting price targets as high as $200 but many maintaining Neutral or Hold ratings due to execution and capital‑intensity concerns. On several discounted cash‑flow (DCF) models, Intel’s intrinsic value is calculated in the mid‑80s to low‑90s per share range, only slightly above or below the current market price, implying the stock is close to fairly valued on conservative cash‑flow assumptions. Sales‑Based Models Still See Undervaluation Separate from the more conservative DCF work, some valuation frameworks focused on price‑to‑sales (P/S) multiples argue that Intel’s AI‑driven mix and size justify a richer multiple than the market is currently assigning. One such model derives a “fair” P/S ratio of about 15.1x for Intel, compared with an observed multiple closer to 13.1x at the time of analysis, suggesting the stock trades at a discount to what its AI exposure and margin profile would warrant. Another narrative points to a fair P/S ratio nearer 17.9x , versus a contemporaneous multiple around 7.6x. Under that lens, Intel looks significantly undervalued on sales even if cash‑flow‑based intrinsic value appears only modestly above the share price. These sales‑centric approaches underpin much of the “still cheap” messaging, emphasizing Intel’s potential rerating as AI revenue becomes a larger and more stable component of the business. Not All Analysts Buy the Undervaluation Story Despite the enthusiasm around AI, some research houses remain skeptical that current valuations can be justified. Early in 2026, one widely cited report called Intel “overpriced” and warned that the shares were trading more than 30% above a fair value estimate of $32 per share , based on cautious assumptions about profitability and competitive risks. Although that figure has since been raised substantially by the same provider, the earlier stance illustrates how sensitive Intel’s perceived fair value is to underlying assumptions about AI demand durability, manufacturing execution and capital allocation. Even after upgrading their models to reflect the AI boom, some analysts argue that Intel’s stock has already priced in a great deal of optimism and may struggle if AI infrastructure spending normalizes or if rivals capture outsized share of accelerator and server CPU markets. AI Capital Raise Adds Another Layer to the Debate The valuation controversy has been sharpened by Intel’s recent decision to raise a large amount of equity capital to fund its AI ambitions. In mid‑August, the company launched a stock offering initially sized at $15 billion and then expanded it to $20 billion after strong investor demand. The sale briefly pressured the share price but was interpreted by some market watchers as a sign of management’s confidence in the scale of Intel’s AI opportunity and its foundry road map. For bullish valuation frameworks, the capital raise is seen as necessary fuel for growth; for skeptics, it reinforces concerns about dilution and the high cost of competing at the cutting edge of semiconductor manufacturing. A Wide Valuation Range, Driven by AI Assumptions As of late August 2026, Intel’s fair value estimates span a remarkably wide range — from the $80–$120 band common among traditional analysts to user‑driven narratives north of $500 per share. The claim that Intel could be roughly 82% below fair value relies on the most optimistic of these models, which assume sustained AI‑powered growth and premium valuation multiples over many years. For investors, the gap underscores how pivotal AI is to the Intel story: the more confidence markets place in Intel’s ability to convert its early AI momentum into durable, high‑margin earnings streams, the more plausible the higher end of that valuation spectrum becomes.

Nic Reeve·14:00 28.08.2026

AI & Tech

AInews showdown: Gemini Omni Flash or Kling 3.0 for next‑gen video control?

AInews showdown: Gemini Omni Flash or Kling 3.0 for next‑gen video control? On 19 May 2026, Google began rolling out Gemini Omni Flash to its apps and APIs for conversational video generation and editing, while Kuaishou’s Kling 3.0, launched on 4–5 February 2026, pushed multi‑shot continuity and cinematic storyboards into the mainstream AInews race. What exactly are Gemini Omni Flash and Kling 3.0? Gemini Omni Flash is Google’s fast, text‑and‑image‑to‑video model built for interactive editing, now generally available as Gemini Omni 1.1 Flash. Kling 3.0 is Kuaishou’s third‑generation family of multimodal video and image models, centered on the Video 3.0 and Video 3.0 Omni engines for short, native 4K clips with storyboard controls. Both systems sit at the frontier of AI video. According to Google’s Gemini model announcement in May 2026, Gemini Omni Flash turns text prompts and optional reference images into short video clips and lets users "easily edit your videos through conversation" in the Gemini app, Google Flow and YouTube tools. According to Google’s developer documentation, the Gemini Omni Flash API is described as a video "generation and editing" model that refines clips via natural‑language conversations and supports video extension. According to Google’s August 27, 2026 release notes, Gemini Omni 1.1 Flash has reached general availability and replaces the earlier preview endpoint, which will be deprecated on September 30, 2026. According to AI Wiki’s Kling 3.0 entry updated September 11, 2026, Kling 3.0 includes Video 3.0, Video 3.0 Omni, Image 3.0 and Image 3.0 Omni, all built on a unified multimodal architecture that outputs short clips with native 4K resolution and synchronized multilingual audio. According to Genra’s February 20, 2026 guide, Kuaishou timed the Kling 3.0 public release to February 5, 2026, with text‑to‑video, image‑to‑video and reference‑driven modes across the lineup. How does Gemini Omni Flash handle video editing and user control? Gemini Omni Flash focuses on conversational editing: users can ask for changes, extend scenes, and adjust frames using natural language, with support for incremental 10‑second extensions up to 40 seconds in Gemini Omni 1.1 Flash. This makes Google’s model feel like an interactive editor rather than a one‑shot generator. Editing in Gemini’s ecosystem is built around back‑and‑forth dialogue. According to Google’s Omni 1.1 Flash blog on August 27, 2026, the model delivers "studio‑quality video production" and lets users extend videos in 10‑second increments, up to a cumulative 40 seconds, while analyzing up to 10 seconds of prior context instead of just the last second. According to the same post, Gemini Omni 1.1 supports features such as first‑and‑last‑frame interpolation and 4K output in supported environments, improving continuity between edits. According to Google’s Gemini Omni product blog from May 19, 2026, users can "easily edit your videos through conversation" and are already seeing the model integrated into the Gemini app, Google Flow and YouTube Shorts, with rollout to Google AI Plus, Pro and Ultra subscribers globally and free use in some YouTube tools. According to the Gemini Omni Flash API documentation, developers can refine and edit generated videos by sending natural‑language instructions in an ongoing interaction, positioning the model as a dynamic editor that supports video extension as well as generation. According to a July 16, 2026 Google Workspace announcement, Gemini Omni Flash now powers Google Vids, where users can edit videos using simple text prompts and generate new clips featuring personal avatars that look and sound like them. According to ilisai’s explainer updated September 2, 2026, the service’s video generator now uses Gemini Omni 1.1 Flash and bills at least 10 seconds per clip, indicating that Google’s fast model is already deployed in third‑party platforms. This conversational workflow favors creators who expect to iterate rapidly, like social video editors or marketing teams that want many small changes without rebuilding clips from scratch. What does Kling 3.0 offer in multi‑shot continuity and storyboarding? Kling 3.0’s Video 3.0 and Video 3.0 Omni models emphasize multi‑shot generation: up to six connected shots in a single clip, with stable character identity, lighting and environment across cuts. Shot planning can be automatic or fully custom, turning prompts into structured mini‑sequences. Multi‑shot tools make Kling feel like a pre‑visualization engine for directors. According to Kling’s Video 3.0 user guide last updated August 26, 2026, the model supports two modes for multi‑shot video: "Multi‑Shot" and "Custom Multi‑Shot". When Multi‑Shot is enabled, it automatically plans transitions and creates multi‑scene content; Custom Multi‑Shot lets users configure shot counts and durations. According to Kling’s July 28, 2026 multi‑shot guide, Multi‑Shot structures a scene through camera coverage, shot changes and narrative progression, reading coverage and shot information from the prompt to adjust angles and compositions for cinematic storytelling. According to Morphic’s August 2026 Kling 3.0 guide, Kling Video 3.0 supports multi‑shot sequences of up to six camera cuts per generation, with text‑to‑video, image‑to‑video and start‑and‑end‑frame‑to‑video modes within a maximum duration of 15 seconds per clip. According to Invideo’s May 28, 2026 overview, Kling 3.0 can generate up to six connected shots while letting users either describe the scene and let the model plan cuts or specify each shot’s framing, duration and camera movement for precise shot‑list execution. According to Kling3Pro’s March 26, 2026 feature page, Kling 3.0 multi‑shot generation defines up to six individual shots inside a single 15‑second clip, each with its own prompt and camera angle, while locking character appearance, wardrobe and environment continuity via scene‑level identity encoding. According to AI Wiki and Synthszr’s product ranking updated September 6, 2026, Kling 3.0’s unified architecture produces native 4K video at up to 60 frames per second and supports multi‑shot storyboards with up to six camera cuts, reinforcing its role in high‑fidelity continuity. These continuity guarantees matter for ad agencies, pre‑viz teams and independent filmmakers that need a sequence of connected shots, not just isolated clips. How do lengths, resolution and audio capabilities compare? Gemini Omni Flash emphasizes flexible duration via extensions and focuses on fast 720p clips in many deployed services, while Kling 3.0 centers on short but dense native 4K sequences up to 15 seconds with synchronized multilingual audio. The technical trade‑offs shift who benefits most from each system. According to Google’s Omni 1.1 Flash blog, users can extend a video by 10‑second increments, up to 40 seconds total, with Omni analyzing up to 10 seconds of prior context to keep motion and composition aligned. According to ilisai’s July 19, 2026 article, the original Gemini Omni Flash preview produced short 720p clips from text prompts or reference images and, as of a September 1 update, every video generated with Gemini Omni 1.1 Flash bills at least 10 seconds of output. According to AI Wiki’s Kling 3.0 profile, the new generation moved from roughly 10‑second 1080p clips in Kling 2.6 to 15‑second native 4K clips in Kling 3.0, adding synchronized lip‑synced audio across five languages. According to Morphic’s technical table, Kling 3.0’s Video 3.0 model supports durations between 3 and 15 seconds, aspect ratios such as 16:9, 9:16 and 1:1, and native 4K resolution with other options at 1080p and 720p. According to Synthszr’s September 6, 2026 ranking, Kling 3.0 natively generates 4K video at up to 60 frames per second with synchronized audio and supports up to six camera cuts per clip. Users chasing maximum resolution and integrated audio will lean toward Kling; teams optimizing for iterative editing inside existing Google tools may accept lower resolution in exchange for speed and integration. Where are these models available and how are they priced? Gemini Omni Flash is woven into Google’s subscription tiers and tools, from the Gemini app to YouTube products and Google Vids, while Kling 3.0 is accessible through Kuaishou’s platforms and partner APIs aimed at creators and developers. Commercial terms vary, but both target professional and prosumer use. According to Google’s May 19, 2026 Gemini Omni launch blog, Gemini Omni Flash started rolling out to Google AI Plus, Pro and Ultra subscribers globally through the Gemini app and Google Flow, and became available at no cost in YouTube Shorts and the YouTube Create app. According to the July 16, 2026 Google Workspace blog, Gemini Omni Flash now powers Google Vids, giving Workspace users access to text‑prompt‑based editing and avatar generation within a productivity suite. According to Gemini API release notes, Gemini Omni 1.1 Flash reached general availability in early September 2026, signaling that production billing and quotas now apply as the preview endpoint approaches deprecation. According to Kuaishou’s February 9, 2026 feature guide, Kling 3.0 was officially launched on February 4, 2026 at 11:00 PM Beijing time, with API access for developers beginning February 5, 2026. According to Genra’s February 20, 2026 overview, Kling 3.0’s rollout prioritized "Ultra" subscribers before opening more broadly, positioning the models as premium tools for serious creators. According to Morphic’s guide, third‑party platforms integrate Kling 3.0’s modes into their own interfaces, offering creators control over duration, resolution and multi‑shot features alongside their own pricing. According to Synthszr’s September 2026 ranking, Kling 3.0 appears in AI product lists targeted at production users, indicating its positioning in professional and semi‑professional video workflows. These distribution strategies matter. Google is tying video AI tightly to its productivity and social stacks, while Kuaishou and its partners push Kling into dedicated creative and editing environments where users may build entire pipelines around it. Who gains more from editing flexibility, and who needs multi‑shot continuity? Creators who iterate quickly on single clips—with frequent text‑driven tweaks, avatar changes and scene extensions—gain most from Gemini Omni Flash’s conversational editing and deep integration in Google tools. Teams planning storyboards or ad sequences benefit more from Kling 3.0’s multi‑shot continuity and 4K, audio‑rich outputs. Different workflows point to different winners. For social managers and short‑form creators inside YouTube and Workspace, Gemini’s ability to extend scenes, interpolate frames and apply natural‑language edits—"make this shot closer," "brighten the background"—reduces friction in turning rough ideas into polished clips. For cinematographers, agencies and pre‑viz teams, Kling’s combination of up to six connected shots, locked character identity and 4K visuals means they can block out miniature storyboards, test camera coverage and maintain continuity shot by shot. According to Invideo’s comparison, Kling 3.0 explicitly contrasts multi‑shot support against single‑shot models, highlighting that it can either auto‑plan coverage or follow a detailed human‑written shot list. According to Google’s Omni 1.1 blog, the extended context window and interpolation tools are framed around "studio‑quality" production for creators who may not want to think in discrete shots but still care about smooth motion and consistent framing in the finished video. No single model wins outright. The choice turns on whether a creative team thinks in clips with conversational edits or in sequences of shots with tight continuity and high‑end visuals.

Nic Reeve·14:00 11.09.2026

OpenAI Dismisses Three Safety Researchers After Confidentiality Investigation | AI News

What did OpenAI announce on October 1?

Who were the researchers?

What information was allegedly shared?

Why does the case matter to AI safety teams?

What remains unknown about the investigation?

What happens next for OpenAI and its safety work?

Sources

Related Articles

AInews Weekly: Education Gaps, Datacenter Surge and Rutgers Trust Study

Intel’s AI Surge Sparks Fierce Valuation Clash as Some Models Flag 80%+ Upside

AInews showdown: Gemini Omni Flash or Kling 3.0 for next‑gen video control?