Artificial intelligence (AI) has become an increasingly important tool in various fields, including cybersecurity and military operations.
However, renowned cybersecurity expert, Josh Lospinoso, has warned of the growing threat of so-called “adversarial AI” in military systems, raising significant concerns for the security and effectiveness of the military.
Data poisoning, a latent threat
Lospinoso stressed the importance of understanding data poisoning as a form of digital misinformation. If adversaries manage to manipulate the data that AI-based technologies process, they could significantly affect how they work.
Although it has not yet been widely observed, Lospinoso mentioned an emblematic case that occurred in 2016 when Microsoft launched the Twitter chatbot called Tay. Malicious users took this opportunity to flood you with abusive and offensive language, leading to the generation of inflammatory content and forcing you to go offline.
AI as a cybersecurity tool
The expert noted that AI is used in various fields of cybersecurity, such as email filters, to detect spam and phishing attacks. However, he also highlighted that offensive hackers use AI to try to defeat these ranking systems, known as “adversarial AI.”
This situation demonstrates the need to continually strengthen security measures to protect military systems and prevent intrusions.
Vulnerabilities in military software systems
Lospinoso referenced a troubling 2018 Government Accountability Office report, which revealed that most newly developed weapons systems posed mission-critical vulnerabilities.
In this regard, he posed two key challenges: adequately securing existing weapons systems and addressing the security of artificial intelligence algorithms that could be incorporated.
He warned that many military systems are decades old and have been upgraded with digital technologies, making them attractive targets for cyberattacks once access has been gained.
The headlong rush to AI
Lospinoso expressed concern about the possibility of a headlong race to develop and implement AI products. While he stressed the importance of not stopping research in this field so as not to fall behind competitors such as China, he warned against the risks of launching AI products on the market without regard for necessary safety and responsibility.
He stressed the need to accelerate the development of AI without neglecting the fundamental aspects that guarantee its safe use.
The debate on the use of AI in military decision-making
Lospinoso was emphatic in stating that the artificial intelligence algorithms and the data collected are not yet ready to allow a lethal weapons system to make decisions on its own.
He stated that we are far from reaching that level of development and that the risks and ethical implications need to be carefully considered before adopting such an approach in the military arena.
Josh Lospinoso’s warning highlights the importance of addressing “adversarial AI” challenges and vulnerabilities in military software systems.
While AI offers significant opportunities to improve security and efficiency in military operations, it is crucial to maintain a balanced approach that ensures systems protection and responsible decision-making.
The rapid pace of AI development requires constant risk assessment and implementation of appropriate security measures to prevent future threats.
“Adversarial AI” refers to the use of artificial intelligence by offensive hackers to try to defeat AI-based cybersecurity systems.
According to the 2018 Government Accountability Office report, most newly developed weapon systems have mission-critical vulnerabilities, posing a risk to military security.
It is crucial to consider safety and liability when developing and bringing AI products to market. A headlong rush into AI development without ensuring these aspects can have negative consequences.
Although AI has the potential to improve military decision-making, experts note that current algorithms and data are not yet ready to enable lethal weapons systems to make autonomous decisions.
Challenges related to “adversary AI” and vulnerabilities in military software systems must be addressed to ensure the safety and effectiveness of military operations and protect systems against future threats.