AI Industry Development & Policy Dynamics

Gate News message, April 23 — Google's Jules team announced the opening of a waitlist for a new version of the product, repositioning Jules from an asynchronous coding agent to an end-to-end agentic product development platform. According to the official description, the upgraded platform reads enti

Perplexity uses SFT followed by RL with Qwen3.5 models, leveraging a multi-hop QA dataset and rubric checks to boost search accuracy and efficiency, achieving best-in-class FRAMES performance. Abstract: Perplexity's post-training workflow for web-search agents combines supervised fine-tuning (SFT) to enforce instruction-following and language consistency with online reinforcement learning (RL) via the GRPO algorithm. The RL stage uses a proprietary multi-hop verifiable QA dataset and rubric-based conversational data to prevent SFT drift, with reward gating and within-group efficiency penalties. Evaluation shows Qwen3.5-397B-SFT-RL achieving top FRAMES performance, 57.3% accuracy with a single tool call and 73.9% with four calls at $0.02 per query, outperforming GPT-5.4 and Claude Sonnet 4.6 on these metrics. Pricing is API-based and excludes caching.

Gate News message, April 23 — TikTok announced a comprehensive crackdown on AI-generated content that infringes on user rights, disclosing that it has removed over 538,000 videos and penalized more than 4,000 accounts to date. The platform will prioritize enforcement against AI deepfakes, voice

Gate News message, April 23 — Fu Peng, chief economist of Xinfire Group, shared his outlook on the convergence of traditional finance and crypto assets during the 2026 Hong Kong Institutional Digital Wealth Management Summit. According to Fu, the integration of traditional financial institutions wit

OpenClaw switches from Pi to Codex harness to fix a silent authentication fallback, with two PRs addressing the bridge and fallback; post-fix, the agent shifts from shallow heartbeat polling to a full work loop, enabling progress. Abstract: OpenClaw’s Codex harness optimization addressed a critical authentication flaw that caused silent fallback to the Pi harness when using Codex with OpenAI models. Two pull requests fix the authentication bridge and prevent silent fallback, changing the runtime adapter. As a result, agent behavior evolves from shallow heartbeat polling to a full work loop that reads context, analyzes tasks, edits repositories, and verifies progress, improving continuity and visibility across heartbeats.

Bloomberg reported that a private forum group gained unauthorized access to Mythos through legally authorized use by a third-party contractor of Anthropic. Mythos is an enterprise-grade, defensive AI that is only available to large organizations that undergo rigorous review. The group used knowledge of the model URL to guess the system’s location to gain entry, and provided screenshot demonstrations, claiming it was still being used with no malicious intent. Anthropic is investigating and has preliminarily determined it was abuse of permissions rather than external intrusion. This case highlights the risk of entrusting highly sensitive models to third-party oversight, and underscores the need to strengthen governance resilience and trust mechanisms.

According to an alert posted on the X platform on April 23 by SlowMist’s Chief Information Security Officer 23pds, the hacker group ShinyHunters claims it has breached internal systems related to the Anthropic Mythos models and has publicly shared screenshots as evidence, including a user management panel, an AI experiment dashboard, and analyses of model performance and costs. However, Anthropic has not yet issued an official statement.

Chrome Enterprise launches Gemini-powered Auto Browse and Chrome Skills, letting the browser automatically carry out multi-step tasks, but requiring the user to click to confirm; it can save/share AI workflows and integrates with Gmail, Calendar, and Drive, including DLP controls, costs $6 per month, and is positioned as turning the browser into an AI colleague.

OpenAI launched Workspace Agents on April 22 in ChatGPT Business/Enterprise/Edu/Teachers, powered by Codex, designed for long-running cloud operation, shared by teams, and capable of offline execution. They can proactively respond in Slack and generate invoices, execute multi-step workflows, and support scheduling. The research preview is free until May 6; afterward, it will use a credit-based pricing model, with rates to be announced. They compete alongside Google Gemini Enterprise Agent Platform and Anthropic Claude Cowork. All three focus on enterprise-grade agents, but their positioning differs.

Google Cloud unveiled the Gemini Enterprise Agent Platform at Cloud Next 2026, integrating model selection, agent building, DevOps, orchestration, and enterprise security controls, and launched a $750 million fund to help McKinsey, Accenture, and Deloitte deploy enterprise agents. The platform, along with Ironwood TPU, A2A, and MCP, builds its own full-stack offering and consulting channel, to counter OpenAI Operator and Anthropic Claude Enterprise.

According to a report from The Verge on April 22, Anker has officially released its self-developed artificial intelligence chip Thus, positioning it as the world’s first neural network audio AI chip based on an In-Memory Computing architecture. The first batch will be built into Soundcore’s upcoming flagship headphones.

Google announced new security features and deeper integration of Wiz, the Israeli cloud security firm it acquired for US$32 billion, across Google Cloud and rival platforms at its Cloud Next '26 event. The company introduced three AI agents for Security Operations in preview mode, designed for

Microsoft commits AU$25B by 2029 to expand AI and cloud infrastructure in Australia, deepen cyber defense with government agencies, train 3 million in AI by 2028, and coordinate data-center and AI policy to bolster sovereignty. Microsoft’s AU$25 billion expansion in Australia through 2029 aims to grow local AI and cloud capacity, building on a AU$5 billion prior commitment. The plan expands the Cyber Shield program with the Australian Signals Directorate, trains 3 million people in AI by 2028, and partners with the Australian AI Safety Institute, while formalizing data-center and AI infrastructure expectations with the government to bolster digital sovereignty.

Alibaba opens Qwen AI Services to external partners; China Eastern Airlines becomes first outside Alibaba to deploy an end-to-end AI flight booking service, enabling users to request options like affordable direct flights or spacious seats.

Gate News message, April 23 — Australia said it was working with Anthropic on possible cybersecurity flaws after the limited release of Mythos AI model raised concerns about critical infrastructure. The central banks of Australia and New Zealand were monitoring the release, while the government trac

Gate News message, April 23 — Google announced on April 22 that it will release separate eighth-generation TPU chips for training and inference later this year, replacing its previous combined design. The move targets AI agent workloads and offers Google Cloud customers an alternative to Nvidia

Stch, a Bengaluru textile startup, raised $7M to expand its AI-driven fabric reverse-engineering platform and scale partner mills in India and Bangladesh. Abstract: Stch, founded in 2025 by Narahari Payala and Aseem Chitkara, is an asset-light company that uses AI to convert fabric images and descriptions into technical specifications for manufacturing, partnering with mills rather than owning factories. With a $15 million-plus order book from UK, Europe, and US brands, and a pre-series A of $7 million led by Omnivore, the firm aims to broaden its AI capabilities and fabric recipes while diversifying production across regions to mitigate geopolitical risk in fashion manufacturing.

Gate News message, April 23 — Jack Yi, founder of Liquid Capital, shared on X that past crypto VC and project failures were largely driven by teams wasting capital developing unnecessary Web3 products by incorrectly benchmarking against Web2. Yi argued that Web3 is fundamentally a financial

FTX's unliquidated stake values would total about $114B today, with Anthropic $82.3B, SpaceX $15B, Solana $5.1B, Cursor $3B, Robinhood $4.9B, and Genesis Digital $3.5B, per Cointelegraph. Abstract: If the FTX bankruptcy estate had retained its stakes, it would hold roughly $114 billion in assets, led by Anthropic and SpaceX, highlighting large unrealized gains across the portfolio, per Cointelegraph.

US Commerce Secretary Howard Lutnick stated on April 22 that Nvidia has not sold its H200 AI chips to Chinese customers, citing lack of Beijing approval and China's prioritization of investment in domestic suppliers, according to Reuters. The Trump administration cleared some China-bound H200