Who or What is Anthropic?

Anthropic is research company working on artificial intelligence. It was founded by and currently run by former executives of OpenAI. Its legal status is as a public benefit corporation aimed at creating AI systems that are not only useful but inherently safe for both the user and the public in general. Ideally this company is following Isaac Asimov’s “laws of robotics” in which artificial systems are designed not to directly harm humans or by neglect allow harm to happen. The company has released the “Claude” series of chatbots which can code, reason and engage in deep analysis using what they refer to as Constitutional AI and always adhere to a basic set of safety principles.

Who Uses or Works With Anthropic?

The status of Anthropic is such that it partners with Palantir, Amazon Web Services, and Google Cloud. It is the only AI model that the US military uses for classified missions. The company appears to follow its promise to put safety first instead of rapidly turning out as much software as it can in search of profits! Thus Anthropic appears to following up on its promise to work to benefit humanity as opposed to always putting profit first.

Where Is the Danger in Anthropic’s New AI Software?

The New York Times reported that banks have been warned about the new Anthropic AI software and are being given time to upgrade their legacy software in preparation for wider release of this AI tool or similar technology. While non other than the Secretary of the Treasury has been involved in warning banks and other institutions it would appear that the crypto world needs to get ready as well.

AI Risks to the Crypto World

Specific risks posed to crypto systems by increasingly smart and powerful AI tools include the ability to attack and take funds from individual’s crypto wallets when they are “hot” or actively connected to the internet, being able to rapidly spot weaknesses in DeFi projects or contracts and take advantage before those involved in the contracts are even aware of a problem. Additionally, AI tools are increasingly able to generate and insert malware into crypto systems previously thought to be safe. AI will increasingly be able to generate “phishing” schemes across a wide range of users including cryptocurrency users. To the extent that folks are using their own AI tools to develop code for various purposes, more powerful AI programs are able to exploit the code to the benefit of a malicious person and to the harm of the honest AI user.

How Can I Protect Myself from AI Crypto Attacks?

Being aware of cyber risks is the first step in being safe. For example, do not leave a wallet in a “hot” state when you are not actively involved in a transaction. When something seems almost too good to be true, as in phishing scams, remember that said thing is probably not true and don’t fall for it. When you are trusting a new AI tool to write code make sure to review the code yourself to make certain that it makes sense. As we often suggest with online trading, only work in simulation until you routinely generate “paper profits.” Apply the same reasoning to any AI generated code. Find a safe way to run it again and again to make certain that it will work for you and not against you. Using viral scans on your own AI generate code is not so bad an idea either. And not putting all of your eggs in one basket is a good idea too. Diversify your crypto holdings to avoid being wiped out by one attack by a bad actor.