AI agents

Citrini Research published a scenario describing how agentic AI could cause mass economic damage over the next two years. In the imagined report, unemployment doubles and the stock market loses more than a third of its value, driven by a negative feedback loop of AI-driven white-collar layoffs and margin pressure. The piece is presented as a scenario rather than a prediction.

The US National Institute of Standards and Technology (NIST) announced the AI Agent Standards Initiative, a listening exercise to build a roadmap for agentic AI, to be run under the Center for AI Standards and Innovation (CAISI). Its scope extends beyond security to interoperability, international standards leadership and open-source agent development. NIST has issued a request for information on agentic AI threats, safeguards and assessment methods, with input due by March 9, and CAISI will hold listening sessions in April.

Reload, an AI workforce management platform, announced its first product, Epic, alongside a $2.275 million round led by Anthemis. Epic runs as an extension in AI-assisted code editors such as Cursor and Windsurf and maintains shared project context, including requirements, data models and API specifications, across coding agents.

Dylan Cosper, host of the Infosys Knowledge Institute Podcast, interviews Prasad Banala, director of software engineering at a large US-based retail organization. They discuss operationalizing agentic AI across the software development lifecycle, including requirements validation, test case generation and analysis, and issue resolution, with strict governance, human-in-the-loop review, and measurable quality outcomes.

San Francisco-based Kana has emerged from stealth with a suite of AI agents for marketing tasks such as data analysis, audience targeting, campaign management, customer engagement, media planning, and optimizing for AI chatbots. The startup raised $15 million in a seed round led by Mayfield. Its co-founders, Tom Chavez and Vivek Vaidya, previously built Rapt, Krux, and the startup studio super{set}, where Kana was incubated for nine months.

Alibaba released Qwen3.5, its newest AI model series, in an open-weight version that users can download, run, fine-tune and deploy on their own infrastructure, plus a hosted version called Qwen-3.5-Plus on its Model Studio cloud platform. The open-weight model has 397 billion parameters and supports 201 languages and dialects, up from 82 in the previous generation. Alibaba says both versions match leading models from OpenAI, Anthropic and Google DeepMind, but CNBC could not independently verify those self-reported benchmark claims.

Infosys announced a partnership with Anthropic to build enterprise-grade AI agents, integrating Claude models into its Topaz AI platform to automate complex workflows in banking, telecoms, and manufacturing. The tie-up was announced at India's AI Impact Summit in New Delhi, and Infosys plans to use Claude Code internally to write, test, and debug code. The deal did not disclose a deployment timeline or financial terms.

An opinion-style article argues that by late 2025 enterprise AI had shifted from chat-based LLMs to autonomous agents, driven by the failure of standard RAG systems. It states that 72% to 80% of enterprise RAG implementations underperform or fail within their first year, and that agentic RAG introduces a new risk: autonomous execution of malicious instructions. The source text is cut off before its security analysis.

Sam Altman announced on X that Peter Steinberger, creator of the AI agent OpenClaw, is joining OpenAI. Altman said Steinberger has ideas about getting AI agents to interact with each other and that this capability will quickly become core to OpenAI's product offerings.

Margaret-Anne Storey's piece introduces cognitive debt, the understanding lost in developers' heads when they move fast with AI agents, as distinct from technical debt in the code itself. She describes a student team that could no longer make simple changes because no one could explain the design decisions or how the system's parts fit together. The author reports experiencing the same loss on her own projects built by prompting features into existence without reviewing them.

The article argues that as AI deployments grow and packs of agents work autonomously in concert, organizations face a naturally amplified attack surface. It frames multi-agent "swarm" systems as adding security complexity.

Z.ai has released GLM-5, a 754B-parameter model under the MIT license, available on Hugging Face at 1.51TB. That is about twice the size of GLM-4.7, which had 368B parameters and 717GB. The post also notes Z.ai's framing of professional software engineers building with LLMs as "Agentic Engineering," a term the author has seen from Andrej Karpathy and Addy Osmani.