Google launches cross-vendor AI watermark detection, enabling checks for ChatGPT-generated images too.

Beating AI News: Perplexity has open-sourced two multimodal embedding models, pplx-embed-v2-late, with sizes of 0.6B and 9B parameters respectively. The two models are vector-compatible, allowing developers to use the larger 9B model for building knowledge bases while leveraging the smaller 0.6B model for daily searches—eliminating the need to run the large model for every query. The new models support text, image, and PDF page retrieval. Traditional embedding models typically compress content into a single vector, which often leads to detail loss. In contrast, the new models retain a 128-dimensional vector for each token, enabling search queries to match relevant segments within documents directly. When processing PDFs, PPTs, and scanned materials, the models can convert page images straight to vectors, bypassing OCR text extraction while preserving charts, tables, and layout information. Both models are trained from the same 18B teacher model and share a common vector space. In official ViDoRe v3 image retrieval tests, the 0.6B model scored 62.3% when used for both database building and queries; switching to the 9B model for database building and the 0.6B model for queries boosted the score to 63.5%; the full 9B model achieved 65.2%. The model weights are now available on Hugging Face under the MIT license.

Regarding some payment platforms that are registered solely as fintech companies without holding any financial licenses, Hong Kong’s Secretary for Financial Services and the Treasury Chan Ho-lim said that appropriate enforcement actions will be taken as needed under the relevant statutory regulatory framework to maintain financial stability and safeguard the rights and interests of customers and the public. Under Hong Kong’s Payment Systems and Stored Value Facilities Ordinance, it is illegal for any individual or entity to issue or operate stored value facilities in Hong Kong without a license, unless a statutory exemption applies. If cases of suspected unlicensed operation or violations of the Payment Ordinance are identified, the Hong Kong Monetary Authority (HKMA) will directly intervene to handle them, collaborate with other relevant regulators based on the nature of the case, and refer cases to law enforcement departments when necessary.

Chainlink has launched a new DeFi solution, CCIP Vault Adapters, enabling DeFi vaults to directly accept user deposits from over 80 blockchains without requiring vault reconstruction on each chain—significantly expanding access to capital, users, and liquidity. The product is now live, with protocols including Aave, Lombard, United Stables, Veda, and Venus already adopting it. Vaults are a fast-growing vehicle for packaging digital assets and financial strategies into user-friendly DeFi products, with billions of dollars currently deployed in stablecoins, lending, yield strategies, digital credit, and tokenized assets. Yet as the vault ecosystem expands, issuers face competition from capital fragmentation across chains; high-quality vaults are often isolated on a single network, while potential depositors hold assets on other chains. CCIP Vault Adapters are designed to resolve this structural friction.

Beating AI News Brief: Personal AI assistant startup Tab has announced the end of its stealth development phase, hitting a $300 million valuation. Users don’t need to download a separate app; they can send messages via iMessage or WhatsApp to have Tab shop, book flights, or modify hotel reservations. Tab comes with its own dedicated phone number, integrates with browsers, and has a built-in payment wallet. It can open web pages, fill out forms, call merchants directly, wait for the call to connect, then send the results back to the chat window. When processing payments, Tab first displays the merchant, item, and amount, and only completes the transaction after user confirmation. Founded by three entrepreneurs who began developing Tab only this year, its investors include SV Angel, Valar Ventures, and American Spirit. The startup did not disclose the actual funding amount or specific user count. Competing with similar products like Instinct and Meta Muse, Tab positions privacy and payment security as its core selling points. The company states it will not use user data to train its AI, nor will it directly submit bank card information to the AI for processing. However, Tab is currently in an invite-only beta testing phase, and its wallet and auto-checkout features are not yet fully available.

Strategy announced it will release its third-quarter financial results after the U.S. stock market closes on October 29, 2026, and host a live webcast at 5 p.m. ET (Eastern Time) that day to discuss the results. Approximately two hours after the webcast concludes, a replay will be uploaded to Strategy’s investor relations website, with the live stream also available simultaneously.

According to Iran's Tasnim News Agency, Iranian Foreign Minister Araghchi said the negotiation process remains ongoing, with both sides exchanging information via intermediaries. "We have put forward a proposal dubbed the 'Seven-Day Plan' and have received the U.S. side's feedback on it. We are currently reviewing the U.S. side's comments, and I believe we will respond in the coming days," Araghchi stated. He again emphasized that the negotiation process is still ongoing.