# Project Lightwell 是一項龐大計畫，旨在保護開源軟體免受高度先進的 AI 驅動網路威脅。

*genai, business · news · 2026-05-29 · The Financial Express*

## Key points

- Anthropic 即將推出的 Claude Mythos AI 模型能以前所未有的速度識別深層軟體漏洞。
- IBM 推出 Project Lightwell 主要是回應 Claude Mythos 能力所帶來的風險。
- 美國主要銀行如 JPMorgan、Goldman Sachs、Morgan Stanley 和 Bank of America 是 Lightwell 的早期採用者。
- IBM 與 Red Hat 正動員全球超過 20,000 名工程師，主動修補開源軟體。
- Project Lightwell 針對軟體修補，這是大多數現有網路安全廠商未直接處理的領域。

Project Lightwell is a massive undertaking designed to safeguard open-source software from highly advanced, AI-driven cyber threats capable of automated vulnerability exploitation. According to IBM CEO Arvind Krishna, the primary reason behind the multi-billion-dollar investment was the alarming capability of Anthropic’s upcoming next-generation AI model, Mythos. During early evaluation of Claude Mythos, which was released to select infrastructure firms and institutions, the ‘dangerous’ model demonstrated an unprecedented aptitude for locating deep-seated flaws in software code, sending shockwaves through global banking systems and government agencies. “Mythos was the critical triggering factor on this,” Krishna revealed in an interview with CNBC. He added that advanced large language models (LLMs) are “remarkably adept at finding vulnerabilities” and weaponising security gaps in both proprietary and open-source infrastructure. Wall Street names are IBM’s early customers Recognising the severe risk to the global financial system, some of the largest financial institutions in the United States have signed on as early adopters of Project Lightwell. The initial roster includes banking titans like JPMorgan Chase, Goldman Sachs, Morgan Stanley, and Bank of America. “They will use the latest tools to figure out where they might have a vulnerability and where there isn’t a patch that is already available,” Krishna noted. Note that most institutions utilise open-source software, which eventually forms the foundational backbone of modern corporate infrastructure. While many may question the logic behind it, companies choose the open-source platforms due to their accessibility and cost-efficiency. However, the transparent and public-facing architecture also leaves it highly vulnerable to AI-powered scanners that can discover and attack zero-day flaws instantly. With 20,000 engineers, IBM wants to patch the gaps To neutralise fast-moving AI threats before they can be exploited maliciously, IBM is heavily leaning on its open-source subsidiary, Red Hat. Both tech giants are mobilising a massive global force of more than 20,000 software engineers dedicated to helping corporate and institutional partners secure and patch their active software pipelines. Despite entering the cybersecurity arena with significant financial backing, Krishna clarified that IBM does not view legacy or incumbent cybersecurity firms as rivals. Rather, he envisions Project Lightwell as an essential layer addressing a critical gap in the current market. “They’re great at protecting the perimeter, they’re great at figuring out what’s going on, but they don’t do patching, and they don’t do the protection of other software,” Krishna explained regarding traditional cybersecurity vendors. “So this, I think, is a great complement to what they do.”

**Companies:** IBM, JPMorgan Chase, Goldman Sachs, Morgan Stanley, Bank of America
**Countries:** United States

[Read the full story on The Financial Express](https://www.financialexpress.com/life/technology-ibms-5-billion-bet-on-ai-cybersecurity-project-ceo-says-mythos-was-the-triggering-factor-4254202/)

---

Canonical: https://newsio.io/zh-TW/n/d7973cde-2693-400c-909f-ef30868e6d9d/project-lightwell-ai
Summarized by Newsio from The Financial Express. https://newsio.io/how-it-works