In Roman history, Julius Caesar was ordered by the democratic Senate of the Roman Republic not to cross the Rubicon with his army. The Rubicon was a shallow river in northern Italy that marked the boundary between the province of Gaul and Italy. Under Roman law, generals were forbidden from crossing the river into Italy with their armies, as doing so was considered a declaration of war against the Roman Republic.
In 49 BC, Caesar defied the Roman government and crossed the Rubicon, prompting the Roman Senate to flee to Greece. The move gave Caesar control of Rome and sparked the civil war that transformed the Roman Republic into the Roman Empire.
The phrase “crossing the Rubicon” has since come to signify a pivotal point of no return—a tipping point in history that reshapes political and military power and creates a new narrative.
In 49 BC, Caesar’s army dominated much of Europe because of its superior steel gladius swords and military tactics. By contrast, military dominance in 2026 requires superiority not only on the battlefield but also in cyberspace and digital infrastructure to manage soldiers, equipment and operations.
I interpret the modern-day “crossing of the Rubicon” as the point of no return at which democracies are threatened by the use of artificial intelligence to produce weapons of terror or undermine democratic institutions, much as Rome’s democracy was destroyed after Caesar crossed the Rubicon.
In my view, AI has now crossed that metaphorical Rubicon because of its potential to become a weapon in the wrong hands. Major military powers such as the United States and China already use AI developed for their own strategic purposes, although access to those systems has generally remained tightly controlled.
AI may have crossed the metaphorical Rubicon with the public release of Anthropic’s Fable 5 model. What makes Fable 5 significant is that it is reportedly derived from the Mythos model used by the US military and the NSA for offensive and defensive cyber warfare operations.
Anthropic released Fable 5 on June 9, acknowledging from the outset that the technology could cause real harm if misused. It did not take long for those concerns to be tested.
Within days, one researcher reportedly bypassed the model’s safety controls, while researchers at Amazon uncovered prompts capable of extracting information that could potentially aid cyberattacks.
Washington responded swiftly. On June 12, the United States ordered Anthropic to suspend foreign access to Fable 5. Unable to verify users’ identities in real time, the company temporarily shut the model down for everyone.
The restrictions were lifted on June 30, and Fable 5 returned on July 1 with stronger safeguards.
It is important to note that Fable 5 was never shown to have carried out a cyberattack. However, given that the US military reportedly uses a similar model, Mythos, for cyber warfare, authorities considered the restrictions necessary to mitigate potential national security risks.
Fable 5 reportedly outperforms OpenAI’s ChatGPT 5.5 and Google’s Gemini 3.1 on complex tasks. Its ability to write computer code and identify software vulnerabilities has made it a potential tool for cyberattacks.
The threat is not confined to cyberspace. In theory, the Mythos model could also be used to synthesise research on chemical or biological weapons, potentially helping a well-resourced organisation develop weapons of mass destruction using information already available on the internet. Fortunately, Fable 5 includes safeguards designed to prevent such misuse.
Even with those guardrails, Fable 5’s advanced coding capabilities could still provide logistical support to militaries and non-state actors. Beyond weapons development, armed forces depend heavily on logistics, management systems and IT infrastructure.
Historically, the first large-scale military deployment of artificial intelligence occurred during the 1990 Gulf War, when the United States used a logistics analysis tool known as DART. That system relied on much older symbolic AI algorithms rather than the neural networks that underpin today’s large language models.
A modern large language model can generate code for a logistics analysis tool similar to DART with relative ease and assist military organisations in implementing systems to streamline their operations.
There is also the growing risk that AI could be used to undermine democracy through the spread of misinformation. AI-powered bots can flood social media with falsehoods, while AI-generated videos and fabricated news reports can be used to manipulate public opinion and influence voters.
It is therefore critical that lawmakers in Trinidad and Tobago receive regular briefings on developments in artificial intelligence and its potential applications.
On the positive side, a powerful model such as Fable 5 could help digitise entire government ministries within months. On the other hand, Trinidad and Tobago is not adequately prepared for the cybersecurity threats that increasingly sophisticated AI systems are likely to pose.
The absence of comprehensive cybercrime legislation and a well-equipped cybercrime unit leaves the country vulnerable to malicious actors seeking to weaponise artificial intelligence.
