The evolution of artificial intelligence (AI) models, such as the one recently announced by the firm Anthropic, is transforming the global cybersecurity landscape. This is because it will allow both specialized hackers, as well as people without advanced programming knowledge, to be able to execute sophisticated attacks against the entire computing ecosystem that governs the world.
This is reflected by Professor Robinson Rivas, from the School of Computing of the Central University of Venezuela (UCV), who warns that platforms such as Fable 5, the most recent innovation from the AI firm Anthropic, They will facilitate the detection of gaps in the global technological infrastructure.
“The new tools (…) what they are going to do is create more gaps and vulnerabilities in the entire global computing infrastructure, including, of course, cryptocurrencies,” says Rivas.
In this scenario, bitcoin (BTC) and cryptocurrency exchanges emerge as the main target of cybercriminals due to the custody of confidential data.
Rivas explains, in dialogue with CriptoNoticias, that, unlike the underlying Bitcoin protocol, whose blockchain has demonstrated mathematical and technological robustness capable of resisting direct attacks, the weak points of the ecosystem are found in financial services companies that manage funds centrally.
As the academic explains, “most of the cryptocurrency infrastructure for users is in business-type wallets, like Binance or Poloniex or any of those large sites. And those sites have a centralized infrastructure that connects to the blockchain. So, it is on those centralized infrastructure sites where attacks using this type of tools will surely focus a lot.
For Rivas, the focus of the computer attackers will not be on altering the distributed registry, but on violating the servers of the exchange houses. The UCV specialist details that “many of these sites store users’ private keys in their infrastructure. And if you manage to steal or access the private keys of the users in the infrastructure, with that you can access the blockchain without any problem.
This attack pattern evokes historical incidents in the industry, to which the professor recalls the attack on Mt. Gox in Japan and other large platforms, such as Bybit in 2025 and Binance’s BNB Chain network in 2022.
“They [los hackers] They have attacked crypto-based banking sites and not the blockchain directly. And we are surely going to see an increase in these types of attacks,” warns the also master in distributed systems.
Rivas’ comments come after Anthropic will release the most advanced AI on the market. This is Claude Fable 5, presented on June 9, and is the first model of the Mythos class available for general use. Anthropic noted that the model exceeds in capacity all those it has made publicly available to date, as reported by CriptoNoticias.
The ineffectiveness of filters and conversational deception
The current debate surrounding the security of tools like Fable 5 or the restrictive government use models known as Mythos largely revolves around the ethical barriers implemented by their creators.
Although firms like Anthropic restrict cybersecurity consultations, Rivas assures that “there has already been evidence that they have always found some way to fool the artificial intelligence filters to be able to access things that were prohibited. “That has happened over and over again.”
Attackers, says Rivas, manage to bypass blocks by disguising intentions within the flow of the conversation with the interactive digital asset. Rivas comments that “it is very likely that, with a series of questions, gaining their trust, the system will be able to give answers thinking that they are not to violate cybersecurity.”
Likewise, it explains that the use of social engineering directed towards legitimate operators represents a direct access route that neutralizes any technical filter. The professor highlights that if access to a model is limited to a state institution, “another way to access that information is for you to hack or do social engineering on the personnel who are using the tool.”
This systemic vulnerability demonstrates, according to the expert, that “none of these platforms is perfect and the only thing that is useful here is permanent surveillance.”
State surveillance and the urgency of global regulatory frameworks
The massification of these tools not only compromises cryptocurrency funds, but also privacy and digital identity management at a global level. Rivas points out that “the sad reality is that we are already at that point where all the state information agencies of all countries, large and small, carry out surveillance on their population.”
“There is not the slightest doubt about that, there is too much evidence,” he emphasizes. Given this, it highlights the importance of civil society pushing for rigorous data protection laws, similar to those of the European Union or the United Kingdom.
Indeed, this concern about mass surveillance and the leaking of confidential data is based on precedents of great international impact. An emblematic case was the PRISM programrevealed in 2013, which demonstrated how intelligence agencies directly accessed to the servers of the main technology corporations to collect communications from citizens on a global scale.
Likewise, the scandal of Cambridge Analytics evidenced the use of social engineering techniques and unauthorized collection of psychological profiles to manipulate political processeswhich demonstrates the vulnerability of digital identity to automated analysis tools.
Now, instead of proposing a stoppage in the development of artificial intelligence, a measure that he describes as useless, he urges regulating the responsibility of technological corporations. “Stopping the development of these things makes no sense,” says the specialist.
Rivas maintains that There must be a “very firm and clear” regulatory framework about the civil and criminal liability of AI companies, arguing that “they are no longer innocent tools.”
“If a third party uses these tools and these tools are not carefully designed, it could be possible that these companies also have liabilities,” he warns.
The final objective, says Robinson Rivas, must be a controlled growth of technology that respects human rights (HR), drawing a historical parallel with atomic energy.
«The great powers did not stop nuclear development, none of them stopped nuclear development, but they reached agreements and civil society and pressure from countries managed to contain these things from being used in a more or less responsible way, limiting the use and limiting the growth of the technology in some way within some parameters. I think that is the moment that happens in artificial intelligence. I repeat, not to stop or censor, but so that there is a much more controlled form of growth,” he concludes.