How to Compare 10-K Risk Factors Year Over Year

If you're an agent: Set up Valyu and make your first search.

If you're a human: Create an account and get an API key.

MCP onboarding: Connect your agent to https://mcp.valyu.ai/mcp and follow the MCP setup guide for authentication.

What will the comparison show?

Suppose you maintain a risk watchlist for Netflix. Item 1A runs to tens of thousands of characters each year, and most of it barely moves. You need to know what actually changed and which of those changes still matter, so you ask:

What changed in Netflix's risk factors between its last two 10-Ks, and has anything newly added since been resolved?

The useful answer is a table of changes with the before and after wording, plus an update from later filings. We will use these documents:

Step 1: define the filings and the change labels

A vague request such as "what are the new risks?" invites a summary. Be specific about:

• The company and both fiscal years, using original filings rather than amendments unless you want those too.
• Topic alignment, so a risk factor that moved is not reported as both removed and new.
• The labels: new, expanded, narrowed, removed and unchanged.
• Short before and after passages for every claimed change, with a source for each.
• A follow-up on filings made after the latest 10-K. Without this instruction, you only learn what was disclosed in January.

Step 2: run the comparison with Valyu

One of the ways to run it is with Valyu connected to Claude or ChatGPT. Follow the chat apps guide to connect it, then ask:

Use Valyu DeepResearch in fast mode to compare Item 1A in Netflix's 10-Ks for fiscal 2024 and fiscal 2025. Align matching topics, label each change as new, expanded, narrowed, removed or unchanged, and give short before and after passages with sources. Then check later Netflix filings and say whether any newly added risk has since been resolved or changed. Ask me to confirm before you start.

You can also run it from your own application with the SDK. Install it with pip install valyu for Python or npm install valyu-js for TypeScript, and set VALYU_API_KEY in your server environment. The Python SDK and TypeScript SDK docs cover the options. Both examples send the same request:

from valyu import Valyu
 
client = Valyu()
 
task = client.deepresearch.create(
    query=(
        "Compare Item 1A Risk Factors in Netflix Forms 10-K for fiscal 2024 and "
        "fiscal 2025. Align matching topics despite heading or order changes. "
        "Identify new topics, expanded or narrowed wording, removed passages and "
        "unchanged topics. For each claimed change provide short before and after "
        "passages and both filing URLs or section references. Do not infer that a "
        "described risk occurred. Mark incomplete extraction rather than treating "
        "missing text as deletion. Then check Netflix filings made after the fiscal "
        "2025 10-K and say whether any newly added risk has since been resolved or "
        "changed."
    ),
    mode="fast",
    research_strategy=(
        "Use the two original SEC annual filings and full relevant risk-factor "
        "sections where retrievable. Verify filing year, entity and section "
        "completeness before comparing. Check whether a passage moved before "
        "labeling it removed. Read later 10-Q and 8-K filings for updates."
    ),
    report_format=(
        "A concise evidence brief and comparison table. Cite each material finding, "
        "retain missing information and separate direct disclosure from "
        "interpretation."
    ),
)
if not task.success or not task.deepresearch_id:
    raise RuntimeError("No task ID returned; inspect the submission before retrying")
 
print("Save this task ID:", task.deepresearch_id)
result = client.deepresearch.wait(
    task.deepresearch_id, poll_interval=5, max_wait_time=900
)
if result.status != "completed":
    raise RuntimeError("Research has not completed")
 
print(result.output)

Check out all of the other integration options here: https://docs.valyu.ai

Step 3: read what changed

The report named one new topic, the proposed Warner Bros. Discovery (WBD) deal, and listed the wording changes around it. These examples are from that report, and we confirmed each one in the filings:

The update on later filings mattered most. WBD ended the merger agreement on February 27, 2026. Netflix's Q1 10-Q says Paramount Skydance, on WBD's behalf, "paid a $2.8 billion termination fee owed to Netflix", and that the related financing arrangements were terminated with no amounts borrowed.

Step 4: check the report against the filings

To test the report, we extracted Item 1A from both filings on EDGAR and ran our own sentence-level text comparison. The core findings matched, and the check turned up three limits worth planning for:

• A smaller change was missed. The fiscal 2025 filing adds that the EU committed in July 2025 "not to adopt or maintain" network usage fees. The report did not list it.
• Two filings were not retrieved. The report said it could not find the Q1 and Q2 10-Qs and flagged that as a coverage gap, not evidence of no change. We read both on EDGAR, and each says there were no material changes to risk factors.
• The unchanged list was harder to verify. Several entries in that section had incomplete citations, so treat "unchanged" as a starting point and spot-check anything you plan to rely on.

When you review your own report, open the cited filings and confirm the entity, fiscal years and section boundaries. Make sure formatting edits and moved passages are not labeled as substantive changes. Then search the newer filing for any passage marked as removed.

Figure 1. How the WBD disclosure moved across Netflix's filings, using figures as reported.

How should you interpret a change?

The decision: move the WBD items off your forward-looking risk list and record them as history, including the $5.8 billion and $42.2 billion figures. Treat the $2.8 billion fee as a Q1 2026 one-off when you compare net income across periods.

The next action: follow the changes that may still be live. The fiscal 2025 filing dates the Writers Guild agreement to May 1, 2026 and the SAG-AFTRA and DGA agreements to June 30, 2026. Those dates have passed, so find out what happened. Advertising also gained new wording on targeting and data. Rerun the comparison when the next 10-K arrives.

The outcome: a short watchlist that separates closed items from open ones, with each entry tied to a filing.

How to adapt the comparison

Change the company and fiscal years, and keep the labels and evidence rules. To follow one theme, add it to the query, for example "only risks mentioning tariffs". For a repeatable run with a spreadsheet, the risk-factor changes guide shows how to request a CSV of topics and passages. For other filing work, see searching 10-K filings via API and researching capital structure and covenants.

FAQ

Where are risk factors in a 10-K?

In Part I, Item 1A. Later 10-Qs can update them in Part II, Item 1A, which is where Netflix's "no material changes" statements appear.

Why not just run a text diff?

A diff is a good check, but it is noisy. Ours flagged quote-mark changes, page breaks and small edits such as "generative artificial intelligence" becoming "generative AI". It also can't tell you what happened after the filing. Use it to test the report rather than replace it.

Does a new risk factor mean something bad happened?

No. Risk factors describe what could happen, and companies also add them for legal reasons. Keep text changes separate from claims that an event occurred or a risk became more likely.

Does missing text mean a risk was removed?

Not until you check. It may have moved, or a partial extract may have cut it off. Search the full newer section before you label a passage as removed.

Which DeepResearch mode should I use?

Fast mode handled this two-filing comparison in our test. For several companies or more years, try a deeper mode and check the pricing first.

What happens if waiting times out?

Keep the saved task ID and resume that task. A local timeout does not cancel remote research, and creating the task again can start another paid investigation.

Build a risk-factor watchlist

Compare the two filings, label each change, then check what happened next. Give your agents access to Valyu, read the Valyu docs, explore the finance documentation, or review the SEC filings API comparison.