Can a prompt hack a network? Hedera explains how to protect yourself

• AI accelerates development but does not replace security audits.
• The ease of generating code does not eliminate security risks in financial applications.

What happens when finding a vulnerability in a network is within reach of a simple prompt? Artificial intelligence (AI) is lowering the barriers to detecting software flaws, thereby increasing the pressure on infrastructures that seek to attract banks and companies. For Luis Alfredo Molano Vega, Developer Relations Community Manager of Hashgraph at Hedera, the answer begins with a design decision: prioritizing security from the inception of a network.

<>, he stated in an interview with CriptoNoticias during the Blockchain Summit Latam Colombia 2026.

The discussion becomes relevant following changes in the vulnerability disclosure policy of Bitcoin Core, which includes addressing detectable flaws with automated tools and even simple instructions to language models.

For Molano Vega, security must be considered from the design of a network and not as a feature that is incorporated later. He also highlighted the importance of knowing who operates the nodes and participates in the consensus, especially when companies need to assess the risks of using a distributed infrastructure.

However, the audit of these networks has another difficulty: financial institutions need to verify operations without exposing confidential information.

<>, he explained. The possibility of establishing access permissions and using cryptographic mechanisms, such as zero-knowledge proofs, allows addressing that tension between transparency and privacy.

These tools can demonstrate that an operation meets certain conditions without necessarily revealing the data that supports it. This is a relevant aspect for entities that need to subject their processes to audits but cannot publish sensitive commercial or financial information.

From functional code to a secure financial solution

Tokenization also introduces demands that go beyond representing an asset on a network. Companies must establish who can issue, acquire, and transfer it, in addition to incorporating know-your-customer (KYC) and anti-money laundering (AML) procedures, according to applicable regulations.

But one of the challenges that could gain greater relevance lies in the development of applications through AI.

Molano Vega pointed out that in hackathons, agents capable of identifying a challenge, developing a proposal, and even submitting it automatically on behalf of a user are already being observed. This accelerates project production, although it does not guarantee that they are ready to operate in real environments.

<<What we have to do is ensure that the user experience has the required security, certifications, and everything else that is necessary, including traceability, to know what is happening>>

Luis Alfredo Molano Vega, Hashgraph.

The difference lies in getting an application to work and demonstrating that it can be used safely. A prototype can surpass a technical demonstration, but it still needs code reviews, dependency audits, security tests, and regulatory controls before managing financial operations.

In that scenario, AI does not eliminate the need for human oversight or the responsibilities of those who develop and circulate these solutions. On the contrary, the easier it is to create software, the more important it will be to verify what it does, what information it uses, and what risks it may introduce.