What public companies just told the SEC about AI, read straight from the filings. Every quote links the original document.
Equifax Claims Non-Public Data as Its Agentic AI Moat
The most assertive AI language in this cycle came from Equifax in its third-quarter earnings release. The prepared remarks go well past generic AI adoption language: the CEO frames non-public proprietary data as the foundation of a competitive moat and names agentic AI specifically as the delivery mechanism for new products.
Our scale proprietary data is the foundation of our AI data moat and a big competitive advantage, and we are expanding our capabilities to leverage our unique, non-public data assets with EFX.AI and agentic AI to rapidly deliver higher-performing scores, models, and multi-market products to help our customers grow.
The argument is more defensible than most companies are making: the value of agentic AI here is gated by data that competitors cannot replicate. Equifax is also one of the few companies this period to tie agentic AI directly to a named proprietary product line rather than leaving it as a capability aspiration.
FactSet Annual Report Names Agentic AI as an Offensive Weapon
In its 10-K for the fiscal year ending August 2026, FactSet Research Systems became one of the first financial data companies to call out agentic AI specifically as a tool used by attackers rather than a technology being adopted internally.
Cyber-attacks continue to evolve in sophistication, frequency and volume, including as threat actors exploit generative and agentic AI to discover and exploit vulnerabilities at unprecedented speed and scale...
The language sits in the cyber risk section alongside phishing, ransomware, and denial-of-service attacks. Naming agentic AI as an active offensive vector in a 10-K is a step further than most companies have gone; most filings still treat it as an internal adoption risk rather than an external threat.
Community Banks Start Naming Specific AI Programs and Threats
First United Corp, a Maryland community bank, listed an "Agentic AI transformation program" as a named strategic initiative in its investor presentation this week, alongside a consumer mobile banking overhaul and a check fraud prevention tool. A named program with its own line in an investor deck implies a budget and an owner, not a general aspiration.
FNB Corp, a Pennsylvania regional bank, used more specific language around a threat it is actively preparing for rather than a technology it is adopting.
...our capability to effectively deploy multi-layered defense strategies to mitigate risks associated with evolving artificial intelligence sourced fraud schemes...
The phrase "artificial intelligence sourced fraud schemes" is concrete enough to suggest the bank's risk team has characterized a category of actual incidents, not just a theoretical concern.
LQR House Discloses a ByteDance AI Agreement
In an 8-K, YHC (formerly LQR House, now pivoting to fintech under the Fusion Five subsidiary) disclosed that Fusion Five entered into an agreement in June 2026 with "a ByteDance group company" for AI and quantitative modeling capabilities. The filing does not name the specific ByteDance entity or describe the commercial terms. The arrangement is described as foundational to Fusion Five's AI-assisted research and portfolio analytics platform. Given the ongoing regulatory attention on ByteDance-affiliated entities operating in US financial infrastructure, the disclosure is worth tracking as this filing matures.
First-Time Filers Worth Knowing
Kinetic Seas filed an 8-K describing MaluOS, an enterprise operating system it says manages agentic AI, generative AI, and conventional database applications within a single environment while preserving "data and AI sovereignty." The company supports both frontier cloud models and open-weight local models. Kinetic Seas has not previously appeared in SEC filings with this language, making this a first-time entrant in the agentic AI category.
AIAI Holdings Corp filed an S-1 describing a roll-up model: acquire companies, improve them through proprietary AI, and generate returns from the operating improvements. The AI itself is licensed from Messier 42 LLC, a company controlled by AIAI founder John P. Rochon. That related-party license structure is the first thing any analyst reading this filing should examine before the company prices.