15-Second summary

Reading through dozens of threat reports to understand a campaign or identify emerging TTPs can eat hours of an analyst’s day.

Ask AI—now part of Feedly Threat Intelligence—lets you quickly summarize up to 25 reports in multiple languages, pull out IoCs, TTPs, and other actionable insights, and create intelligence deliverables in formats you can control.

With Ask AI, you can:

• Synthesize up to 25 reports across multiple languages, detect trends, and surface the most relevant threats within minutes.
• Extract actionable data like IoCs, TTPs, CVEs, and malware references—from selected sources, all mapped by Feedly’s Threat Graph for deeper context.
• Generate intelligence deliverables with complete control over sources, citations, and structure, speeding up dissemination while preserving accuracy.

Why not use another LLM?

Feedly AI is CTI-optimized. Its deeper understanding of adversaries, TTPs, CVEs, and IoCs reduces the inaccuracies you might see with generic language models. Plus, in-line citations link your output back to the original reports for transparent verification and traceability.

Ask AI saves substantial time reading and processing threat intelligence reports to help you identify relevant threats and spend more time on analysis.

The challenge of processing open-source intel

There’s too much open-source threat information for teams to stay on top of. You try your best to focus on the most significant threats or known gaps in your defenses. Extracting patterns across multiple articles and enumerating supporting tactics, techniques, and procedures is highly time-consuming and can be error-prone.

Yet that’s what’s expected of CTI Analysts every day. Read the Internet, determine what’s essential, extract data, and turn it into a professional and actionable report.

To address this challenge, Feedly started building AI Models five years ago and tagging threat intelligence objects in articles to process and filter open-source intel more efficiently. These data are correlated in the Feedly Threat Graph, and today, enable the automatic generation of Insights Cards and Dashboards containing rich context.

Introducing Ask AI

Ask AI is Feedly’s next significant time-saving feature. Analysts can synthesize articles, extract themes, indicators, or entities, and create digestible intelligence deliverables in minutes. Ask AI helps you to interact with articles in AI Feeds, Team Boards, or Search results.

Ask AI leverages Generative AI, which we’ve trained on threat intelligence data to provide deep processing on up to 25 selected articles and perform tasks that would otherwise be manual and time-consuming.

Learn how to use Ask AI in our guide.

Synthesize up to 25 reports across multiple languages

Select up to 25 source reports you want to analyze. Then run an Ask AI prompt to evaluate the content and present the findings in the format of your choosing.

Example 1: Analyze a cyber attack

Analyze a cyber attack by selecting multiple reports about the attack in different languages and run an Ask AI Prompt. In this Prompt, we direct Ask AI to construct a workflow of the attack chain. You can also ask it to list key data like the TTPs and IoCs.

Ask AI Prompt

Construct a workflow of the attack chain.

Ask AI Response

Example 2: Geopolitical analysis

Run prompts on articles and threat reports in foreign languages to better understand key topics in global news coverage. You could analyze the impact of a particular geopolitical event from different viewpoints.

Ask AI Prompt

Produce a long-form threat intelligence report on threat actor's recent campaigns that focuses on the geopolitical risk implications. Report should be in English. 

Ask AI Response

Example 3: Multi-language reporting

In addition to synthesizing foreign language articles, Ask AI can also provide outputs in your preferred language to speed up the sharing of key takeaways across your global teams.

Ask AI Prompt

Provide a summary of these threat intelligence reports in french.

Ask AI Response

Extract actionable data like IoCs, TTPs, and CVEs

Some threat intelligence reports are packed with indicators, adversaries, and other threat data. Analyzing these reports can take hours and runs the risk of missing key data buried within the text. Ask AI enables you to perform tedious tasks like extracting entities and objects from articles and putting them into a usable format.

Example 4: Extracting threat data

Synthesize multiple threat intelligence reports and extract threat actors, TTPs, malware, and IoCs into a table that can be analyzed in a spreadsheet or imported into a security tool.

Ask AI Prompt

Create a table with adversary, TTPs, malware, and IOCs.

Ask AI Response

Generate intelligence deliverables with complete control

Creating deliverables like reports or responses to RFIs can consume the bulk of a CTI Analyst’s day. Asking Gen AI tools risks introducing hallucinations, or producing reports that sound like AI wrote them. Feedly’s Ask AI gives you full control over the sources you analyze to preserve accuracy. It can also be instructed to create a well-formatted output, you can save and apply consistently to similar tasks.

Feedly's Threat Intelligence Advisors created suggested prompts to help answer commonly deliverable requests. You can use them as-is, or modify them to your specifications and save for consistent reuse.

Example 5: Creating a response to an executive inquiry

Suppose you are a CTI analyst and asked to analyze the most relevant threat actors targeting your industry (e.g. Finance), and prepare a summary of the biggest threats.

Ask AI Prompt

I work for a mid-sized bank in USA and need to prepare a summary of the greatest threats by adversaries to my organization.
Produce a BLUF.
Then create a breakdown of the top threat actors and a summary of their tactics.

Ask AI Response

Example 6: Create a threat hunt hypothesis

Take the same example above, and assume you are a threat hunter who needs to prioritize and plan their activities for the next week. Your data needs are more specific. You want to know specific procedures used and how to detect and mitigate them.

Ask AI Prompt

<role>Lead threat hunter</role>

<task>Create a table of attack procedures from the provided article as threat hunt hypotheses</task>

<table_structure>
| Procedure | Description | Logs |
|-----------|-------------|------|
| Short title | Detailed description with patterns | Relevant logs and Event IDs |
</table_structure>

<guidelines>
- Provide detailed technical information
- Structure the information according to the provided table structure format
- Include only actionable procedures for threat hunting
- Focus on specific search patterns
- Avoid generic or ambiguous information
- Include citations
</guidelines>

Ask AI Response

Why trust Feedly’s Ask AI

Feedly AI is CTI-optimized. It’s trained on threat data with a deeper understanding of adversaries, TTPs, CVEs, and IoCs to minimize the inaccuracies you might see with generic language models.

Second, we don’t use your prompts or responses for LLM training.

Third, it’s not a chat that can drift away from the original source content. Each prompt and response is independent to preserve accuracy.

Finally, in-line citations link your output to the original reports for transparent verification and traceability.

If you want to learn more, visit our Ask AI guide.

Summary

Ask AI is not a simple bolt-on Generative AI bot. It builds on years of machine learning development, making our AI Feeds deliver the industry’s most specific and relevant content. It’s fine-tuned on threat intelligence content, and works across single or multiple articles in your Feeds, Boards, and Search.

Ask AI enables you to interact with content, synthesize it, extract themes, and create customized intel deliverables in a fraction of the time previously spent. Ask AI can also perform the tedious work of finding specific threat details to help your team detect, prevent, or mitigate threats. Ask AI is a massive time saver, and because it’s grounded in the content you select and includes citations to sources, you can trust (and verify) accuracy.