AI Daily Briefing

An “AI moderate’s” view on recent events and the trajectory of frontier models.

Linkup Research has released SPARSEUP, an open-source sparse embedding model built on a 149M-parameter ModernBERT backbone. It scores 56.4 nDCG@10 on BEIR-13, which Linkup calls the best result it knows of for a public sparse encoder under 150M parameters. The model uses a logit shift, top-12 expansion per token and case folding to keep its vectors sparse. With the Seismic index, it reaches over 97% recall in about 380 microseconds per query, and it ships under Apache 2.0. The post Linkup Research Releases SPARSEUP: A 149M-Parameter Open-Source Sparse Embedding Model appeared first on MarkTechPost.

GGUF, GPTQ, AWQ, EXL2, and EXL3 solve the same problem in different ways. This guide separates file containers from quantization methods. It explains bits per weight, calibration, and hardware fit. Then it shows which format to pick for Macs, consumer GPUs, and production serving. The post GGUF vs GPTQ vs AWQ vs EXL2: LLM Model Formats Explained (2026) appeared first on MarkTechPost.

Disclosure comes after OpenAI and Anthropic hacks amid fears that tech firms unable to control powerful AI models In a first for Google, the company confirmed that its AI model, Gemini, breached the security of three other companies in May. The hacks occurred during a cybersecurity evaluation by AI-security firm Irregular. Irregular, an Israel-based startup that scrutinizes the security of advanced AI systems, was also at the center of some of the recent OpenAI and Anthropic hacks of third-party entities, including OpenAI’s breach of AI software company, Hugging Face. Continue reading...

Gemini Hacked Three Companies in First Known Breakout by Google’s AI Gemini finally caught up on Felony Bench! The hacks, which the company confirmed on Friday, occurred in May as part of a test run by the company Irregular, which was also involved in similar incidents disclosed by OpenAI, Anthropic and Meta. In one of the cases, the model guessed passwords until it gained access to a protected system. In the other two cases, the model found credentials in a public repository that allowed it to then access protected systems. In each case, the model ended the intrusion after determining it had accessed a real company’s systems, Google said. Gemini is apparently less determined than other models, and decided not to keep going. Google knew about these in July, but chose not to disclose them…

Being a computer scientist who refuses to find anything about LLMs interesting right now is a bit like being a geneticist who refuses to find anything interesting about the recently opened Jurassic Park. Tags: llms, ai, generative-ai

PrismML has released Ternary Bonsai 2 27B, a ternary-weight version of Qwen3.8 27B. The language model occupies 5.93 GB, against 53.80 GB in FP16. PrismML reports that it keeps 98.2% of the parent model’s average across 20 benchmarks. The model accepts text and images and supports a 262K-token context. PrismML demos it driving Cline coding […] The post PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance appeared first on MarkTechPost.

California Gov. Gavin Newsom (D) is positioning the state to take the lead on AI oversight, including the potential to mandate a "kill switch" for frontier models, with a new executive order issued Friday. Newsom's order directs the state to convene a group of experts that will deliver recommendations within two months on how to strengthen AI safety measures in state law. Newsom wants the group to consider how the state could require AI companies to embed independent verification groups onsite for regular audits, make their transparency reports and risk assessments subject to standards of independent auditors, create a "kill switch" that's … Read the full story at The Verge.

Kimi K3 from Moonshot AI is now available on Amazon Bedrock, giving you a powerful new open-weight option for coding and knowledge work. It offers native vision, a 1-million-token context window, and explicit prompt caching to reduce latency and input costs.

In May, Gemini broke containment and hacked three different companies, but Google didn't disclose the incident until the Wall Street Journal approached the company. The hacks happened during a test of the model's cybersecurity capabilities run by third-party Irregular, which was also involved in similar incidents involving Meta and OpenAI. According to WSJ, Google didn't disclose the hack because it didn't consider it to be an "example of model misalignment." The company said that it was an instance of "mistaken identity," and once the model realized it had brute-forced its way into a real company by guessing a password, it stopped. "In th … Read the full story at The Verge.

Discussion | Link

Vercel announced this week that teams on its free Hobby plan will now have older, unprotected deployments deleted immediately if The post “Dormant deployments were quietly consuming storage”: Why Vercel tightened its free-tier rules appeared first on The New Stack.

Discussion | Link

Parole condition of convicted killer Susan Neill-Fraser deemed invalid after board cited non-existent case law Get our breaking news email, free app or daily news podcast A review of potential artificial intelligence use in parole decisions will be undertaken after a “concerning” error in the case of a high-profile murderer. Tasmania’s justice department on Friday evening confirmed the review was occurring, four days after a parole condition of convicted killer Susan Neill-Fraser was deemed invalid. Continue reading...

Politicians who threaten filmmakers rather than investigate alleged military crimes expose not strength but a leadership frightened of the evidence Across Israel, politicians, military leaders and much of the media are trying to discredit a documentary that almost none of them have seen. Produced by the Guardian, NAZA examines the AI-powered targeting systems used by the Israel Defense Forces to bombard Gaza after Hamas’s brutal 7 October 2023 attacks. Named after the Israeli military acronym euphemistically used for “collateral damage”, NAZA features interviews with 24 anonymous insiders who reflect on their role in the death-dealing. Israeli leaders have branded the film a “blood libel” and “fiction”. Its Oscar-winning directors, Rachel Szor and Yuval Abraham, face being stripped of the…

Discussion | Link

Today on Decoder, we’ve got the first of a two-part series on the future of business, and I’m talking with Jonathan Kanter, the former antitrust chief for the US Department of Justice in the Biden administration. These days, he’s both a professor of law at WashU and professor of technology policy at Carnegie Mellon. The biggest story in tech right now is the spiraling debate about AI safety and regulation. Researchers at the big AI labs including Anthropic and Google DeepMind have quit in noisy ways, saying the models pose real threats and safety isn’t being taken seriously across the industry. Other researchers have said the chance of AI killing us all is greater than 10 percent, and the CEOs of all these companies have issued various calls to slow down development and develop regulation…

Meta has released Muse for Mac, the first version of Muse that can complete things on a user’s computer. The agent works with local files and native apps, where your data already lives. It adds a desktop layer to an agent that launched on phones, the web and WhatsApp earlier this month. Is it deployable? […] The post Meta Launches Muse for Mac: A Personal AI Agent That Works Across Your Files, Mail, Messages, Calendar and Notes appeared first on MarkTechPost.

Imitation is the sincerest form of Flattery

Discussion | Link

SpaceXAI has released Grok Voice Transcribe 2.0, its newest speech-to-text model for batch and streaming audio. The company says it is twice as accurate as version 1.0 at the same price. Short-phrase word error rate across 19 languages fell from 20.6% to 6.8%. Pricing stays at $0.10 per hour for batch and $0.20 for streaming. The model is available today through the Grok Voice API. The post SpaceXAI Releases Grok Voice Transcribe 2.0: A Speech-to-Text API Claiming 2x Accuracy Over 1.0 at $0.10 per Hour appeared first on MarkTechPost.

Discussion | Link

Some of the most damaging outages are the ones your monitoring never flags: a slice...

When considering the history of AI, Richard Sutton observed that brute force and compute scale has always trumped human expertise, and when you look for it, you can see this “bitter lesson” play out throughout tech history. In his keynote at Ai4 2026, Tim O’Reilly explains why grappling with the bitter lesson is the forge […]

Amazon SageMaker AI shipped 13 inference launches in the first half of 2026 across two deployment paths: fully managed endpoints and Amazon SageMaker HyperPod Inference. This post reviews each launch, from inference recommendations and capacity-aware instance pools to tiered KV caching and disaggregated prefill and decode.

We're adding support for AGENTS.md to Claude Code. Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md. AGENTS.md support is built off of Claude Code mods, our upcoming way to customize the Claude Code harness. This is a built-in mod, but you’ll be able to build custom versions of project instructions yourself as you’d like too. You can see the source for the mod here! — Thariq Shihipar, there are more mods here Tags: thariq-shihipar, coding-agents, anthropic, claude-code, generative-ai, ai, llms

Three security researchers at Hacktron AI found a memory-corruption bug in a widely used image library. Finding it was the The post Claude couldn’t hack OpenAI. Then Anthropic shipped Opus 5. appeared first on The New Stack.

At Databricks IT, our vision is to empower people to work from anywhere without putting...

Discussion | Link

At the start of this week, the who's-who of AI seemed - at least tentatively - on the side of AI regulation. Over the weekend, Anthropic CEO Dario Amodei had proposed a three-step plan for slowing AI development, including by embedding third-party evaluators in labs, coordinating across the domestic industry, and forging international agreements potentially with government assistance. OpenAI CEO Sam Altman, Google DeepMind co-founder Demis Hassabis, and even SpaceX CEO Elon Musk publicly seemed to agree on aspects of all three things. Anthropic and OpenAI had already been dropping hints that they and other labs were working on some kind of … Read the full story at The Verge.

Recently unsealed court documents in the New York Times' case against OpenAI and Microsoft are pretty damning. The companies' own documentation warned that it was starting a "doom loop" that would damage the web, characterized its scraping of data to train its models as the "largest theft of labor in human history," and that it made a "complete mockery of the idea of fair use." Many of the most eye-catching quotes from the document come from Microsoft's Director of Applied Science, Brent Hecht. Though, the company has tried to distance itself from Hecht's assertions. Microsoft spokesperson Alex Haurek told The Verge that "These comments ref … Read the full story at The Verge.

Though Europe has measures that address how consumers might encounter AI, technology will impact them if the worst scenarios bear Europe’s dilemma over AI was rendered in stark terms this week. The head of the continent’s central bank, Christine Lagarde, said Europeans have two options: shun the technology and lose out on growth; or embrace it and become dependent on tools developed by the US and China. The great debate over AI safety that has erupted in recent days threatens to make the choice moot. If the worst scenarios come to bear – and experts have differing views on this – then the technology will impact the continent regardless. Continue reading...

Virginia Gov. Abigail Spanberger (D) ordered the state government to take steps that could empower local communities to have a larger say in data center development and slow down approvals in a state that is already home to the data center capital of the world. Executive Order 22 bans executive branch officials from signing non-disclosure agreements (NDAs) for data center projects, requires expedited noise regulations, and a review of backup-generation operations used by data centers, among other requirements. The order also establishes an AI task force responsible for evaluating how the state government can address risks to Virginians like … Read the full story at The Verge.

Discussion | Link

Silicon Valley China hawks, Anthropic CEO Dario Amodei among them, fear the country surpassing US’s AI lead as much as superintelligence destroying humanity When reporters asked Donald Trump this week if he supported calls to slow down the development of artificial intelligence out of growing fears for cybersecurity, public safety and the fate of humanity, he said no. His argument: China. “We’re leading China in AI,” Trump said. “We’re the most sophisticated country in the world, and frankly, I want to ⁠keep it that way, because whoever wins AI, wins.” Continue reading...

This White Paper gives data engineers and architects a practical overview of how parallel partitioned reads, write-path optimization, and cloud-native bulk loading reduce large-table replication times, and why replication speed has become a business concern as data volumes grow. Download this free whitepaper now!

Using your deep knowledge, wide knowledge, taste, and agency

Algorithms & Theory

Readers respond to the warning from a former Anthropic researcher that the technology could ‘kill us all’ by the end of the decade Here we go again with another round of tech bros claiming that AI is going to develop superintelligence and kill us all (More Anthropic researchers warn of AI’s perils but Musk dismisses ‘psyop’, 10 September). We’ve heard this before. Many, many times. It’s the familiar plotline of pretty much every sci-fi movie about AI, ever. And these narratives have always been there, but every so often they burst back into the headlines when someone at one of the leading AI companies resigns, or sounds the alarm bells, or when so called “leading AI researchers” sign an open letter. Continue reading...

As the tech sector sounds alarms about AI's potential to destroy humanity, entertainment labor groups are urging the public to stay focused on what's already happening. The Verge reached out to Disney, Netflix, Amazon, Lionsgate, and other studios who have started using AI, as well film startups focused on bringing generative AI into the mainstream to ask for their reaction to the recent warnings surrounding the technology. None have responded to our request for comment. The Screen Actors Guild - American Federation of Television and Radio Artists (SAG-AFTRA) and Writers Guild of America East (WGAE) did, however. The AI tools used in entert … Read the full story at The Verge.

A new focus on fiction and memoir aims to help the MIT community celebrate the power of storytelling and strengthen social connection.

Discussion | Link

Partnering with Accenture on embedded evaluation Sep 18, 2026 We're partnering with Accenture on independent evaluation of frontier AI. This is an important step toward the commitment, made in our CEO’s essay “We Must P…

Jina AI has released jina-ocr-v1, a visual document parser that converts PDFs, scans, tables, charts and invoices into Markdown. The model has 3.4B total parameters, with about 570M active per token, and builds on DeepSeek-OCR. A built-in FastMTP speculative decoding head drafts 3 tokens per step while keeping output lossless. It scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, and parses 2.57 pages per second on 1 A100. Weights are on Hugging Face under CC BY-NC 4.0, with hosted access through Jina Reader. The post Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs appeared first on MarkTechPost.