Hey ChatAI, how can you help with OSINT?

ChatGPT offers an excellent first look at how AI can aid some aspects of Open Source Intelligence (OSINT) and intelligence analysis. 

Open Source Intelligence (OSINT) is the process of gathering information from publicly available sources to support decision-making and risk management. With the rise of the internet and the abundance of information available online, OSINT has become an essential tool for organizations and individuals alike. In this digital age, being able to efficiently collect, analyze, and use information is critical to staying informed and making informed decisions. This is where GPT-3, the cutting-edge language model developed by OpenAI, comes in.

INFORMATION GATHERING

GPT-3 is capable of performing a wide range of tasks, including data analysis and information gathering. Its advanced artificial intelligence algorithms and vast database of information make it a valuable tool for OSINT. With GPT-3, you can quickly and easily gather information from various sources, including social media, news articles, and other online content. This can save you time and effort compared to manually searching for information, and it can also help you uncover valuable insights that you might otherwise miss.

GENERATING TEXT

One of the most notable features of GPT-3 is its ability to generate human-like text. This means that it can write reports, create summaries, and even answer questions based on the information it has been fed. This can be incredibly useful in the context of OSINT, as it can help you quickly digest and make sense of large amounts of information. For example, if you need to conduct research on a specific topic, GPT-3 can provide you with a comprehensive overview of the subject, complete with key insights and takeaways.

REAL-TIME MONITORING

Another advantage of using GPT-3 for OSINT is that it can help you stay up-to-date with the latest developments in your area of interest. With its real-time monitoring capabilities, GPT-3 can alert you to new information as it becomes available. This can help you stay ahead of the curve and make informed decisions quickly. Whether you’re looking to stay informed about a specific industry or keep track of global events, GPT-3 can provide you with the information you need, when you need it.

PRESENTATION

Aside from its ability to gather and analyze information, GPT-3 can also help with the presentation of information. With its natural language processing capabilities, it can create reports, summaries, and presentations that are easy to understand and visually appealing. This can be especially useful for organizations that need to present information to stakeholders, as it can help ensure the information is communicated clearly and effectively.

In conclusion, GPT-3 is an incredibly valuable tool for anyone involved in Open Source Intelligence. Its advanced artificial intelligence algorithms, vast database of information, and ability to generate human-like text make it a versatile and powerful tool that can help you quickly and efficiently gather, analyze, and present information. Whether you’re looking to stay informed, make informed decisions, or communicate information to others, GPT-3 is a must-have tool for anyone working in OSINT.

End of blog post.

No, not really. 

AUTHOR: AI

As some of you may have suspected, the text above was written by ChatGPT or GPT3. I used the prompt, “please write a 500-word blog post in a friendly tone about how ChatGPT can help in Open Source Intelligence (OSINT).” ChatGPT responded in less than a minute, and – although there were some repetitions and a lack of specificity – the AI displayed its potential. I left the text raw and unedited to showcase the output. 

CHATGPT: KEY FACTS

Let’s address what ChatGPT is and how it was trained before we dive down into specifics: 

• ChatGPT is an AI language model. This means ChatGTP processes and generates text based on the input it receives. When you ask a question or make a statement, the AI analyzes the words and tries to understand their meaning. Then, it generates a response using a combination of pre-existing knowledge and language patterns learned from the data it was trained on.
• ChatGPT has limited knowledge of events after 2021, as its primary training data stopped that year. However, it can still respond to queries on issues post-2021, albeit with some limitations. For example, it didn’t know about the ongoing battle for Bakhmut in Ukraine. It instead connected my request to explain the battle for Bakhmut to pre-2021 data and used the initial Russian invasion of Ukraine (2014) and the war in Donbas as context. 
• ChatGPT continues to learn. The AI never stops learning. In fact, one of the main reasons behind the current “free” (you pay with your data) and universal access to the ChatGPT is to help train the algorithm. The more data ChatGPT receives, the better the output becomes. 
• ChatGPT is biased (and open about it). Any machine learning algorithm will be subject to the data fed to it and algorithmic bias. This is obviously one of the many risks with AI, as it could easily automatize large-scale disinformation or scam campaigns in an “authentic” way. 
• ChatGPT is just one of many AI-powered chatbots. The recent frenzy about ChatGPT makes it seems like OpenAI is the sole player in the AI-powered chatbot market. But there are others, for example, ChatSonic, Jasper, DeepL Write, and Rytr. 
• Beware of the privacy implications. There are many privacy concerns, from asking for a phone number to utilizing your conversations as training data. Like in the case of other privacy-sensitive applications, the user must do a cost-benefit analysis and devise offset measures (e.g., using a burner number, refraining from sharing private information in conversations, etc.). However, at the time of the writing, ChatGTP does not support landlines, VoIP providers, or Google Voice. OpenAI offers the option to opt out of having your data used to improve their models. According to their F&Qs, all you have to do is email your request and your organization ID to support@openai.com. 

Casual human-AI interaction at the office

DATA PROTECTION AND THE “BLACK BOX” DILEMMA

Since data protection is essential for our readers, we reached out to information security expert Phillipp Mehl to ask about ChatGPT. Here’s what he had to say:

“Generally speaking, there are two sides to the privacy discussion surrounding large language models and implementations like ChatGPT. One side focuses on user input, the data you write or copy-paste into the prompt box, and the data OpenAI collects when using their services and website. The other side is focused on the existing data in the system and how it is being processed and presented. Both can be relevant to OSINT projects and investigations. 

When you use ChatGPT in a professional setting, you likely act within the scope of EU privacy legislation, like the GDPR. If you handle compliance and data protection responsibly, this can lead to an additional bureaucratic burden. At this point, OpenAI does not provide all the necessary information about data processing to properly fulfill your data protection obligations. 

There is personal data in ChatGPT’s dataset, and OpenAI will use the personal data that you enter into a prompt without the appropriate amount of transparency and traceability. You may still face legal and ethical considerations even if you ignore this issue. For example, are you really working with an OSINT mindset if you cannot confirm the sources of ChatGPT’s information?” 

THE POWER OF THE PROMPT

The secret to unlocking the AI’s potential is using the right prompts. You need to be specific about what you want it to do and re-calibrate the output if the result disappoints. A good prompt sets specific requirements: topics, scope, resources, style, and other parameters, depending on the desired output.

You can task ChatGPT to write a 500-word risk assessment on a specific country from a company’s perspective, include security incidents from a given year, and highlight high-risk regions. Analysts will appreciate the ability to instruct the AI to devise key judgments or present the information in bullet points or with headings. Although students beware, OpenAI released an AI-detection tool. 

ChatGPT can also generate google dorks or Twitter search combinations to make your data hunt easier. Another example is asking ChatGPT to retrieve and arrange specific data into a table (although this can be more or less successful). 

These are just some examples… The possibilities are endless, and it is up to your needs and imagination. 

SO CAN IT HELP WITH OSINT? 

ChatGPT offers an excellent first look at how AI can aid some aspects of OSINT and intelligence analysis. 

From my limited experience with ChatGPT, I would say that the AI is most useful in a “keyboard monkey” role. If adequately instructed, depending on the topic, it can generate a decent first draft or help organize your ideas on “paper.” In other words, it can free up the OSINT or intelligence analyst for more creative and cognitive tasks.

Some clear limitations are the post-2021 data gap (likely temporary until knowledge increases), information-retrieval ability (information was often biased, incomplete, and inaccurate), lack of transparency on data protection, and the “black box” dilemma. 

Despite its positives and negatives, the recent ChatGPT frenzy got the community thinking and debating the role of AI-powered chatbots in OSINT and intelligence. In the future, the OSINT community should reflect on the operational requirements of AI and how we can safely and ethically integrate tools such as ChatGPT into our workflow and methodologies. 

Undoubtedly, AI has a place in OSINT, whether for synthetic imagery generation, video translations, or chatbots- perhaps not as a replacement for the intelligence analyst (not for now), but as a force multiplier, augmentor, and assistant. 

So rest assured, ChatGPT won’t steal your job, but an intelligence analyst with ChatGPT might. 

by Vlad Sutea