ISLAMABAD: Pakistan is considering introducing specific legislation to address crimes involving artificial intelligence (AI), Law Minister Azam Nazeer Tarar told Arab News on Thursday, saying existing cybercrime and criminal laws cover some offenses but may need to be strengthened to tackle new forms of criminal activity involving the technology.

The minister’s remarks come amid growing concerns about the use of AI platforms to research or plan potentially dangerous activities, with Pakistani investigators launching criminal investigations into AI-enabled incidents following tipoffs from the US Federal Bureau of Investigation (FBI).

Speaking to Arab News, Tarar said AI was a relatively new technology, and the government was examining whether additional legal provisions were needed to tackle crimes committed with its assistance.

“It’s a recent innovation and if there are benefits [to it], there are challenges as well, and there are pros and cons. We are conscious of this fact,” he said.

“We have examined situations. Our laws by and large cater for that, but if something else is needed, we are examining in that direction and we’ll incorporate the relevant provisions,” he added.

Asked whether the government planned to introduce new laws to deal with AI-assisted crimes, Tarar said legislation must evolve alongside technological change.

“The rest of the world is also working [along] the same lines. I mean, law is not a stagnant thing. It [keeps] improving or it’s an evolving process,” he added.

Tarar said existing cybercrime provisions and general criminal laws could be invoked in some cases involving AI, though the government was considering more specific legislation.

“Our cybercrime provisions or general provisions in the substantive law also deal with these situations because there are some general provisions which can be invoked [to deal with AI-assisted crimes],” he said. “But as far as the very specific provisions dealing with artificial intelligence [are concerned], yes, we are considering [legislating] in that regard.”

The minister’s remarks follow investigations into two recent cases involving the alleged use of AI to plan crimes by Pakistan’s National Cyber Crime Investigation Agency (NCCIA) last month.

In one case, a 17-year-old student was arrested in Bahawalpur over an alleged plot to poison his father after an FBI alert flagged concerning activity on Anthropic’s Claude AI platform.

According to a first information report registered by the NCCIA on Sept. 23 and seen by Arab News, the alert concerned a user who had conducted “highly concerning, multi-session research regarding the lethality, timing, and optimization of lethal substances.”

The FIR alleges that the teenager attempted to poison his father using laboratory-grade methanol and later planned another attempt involving abrin, a toxin derived from rosary peas.

An NCCIA official told Arab News that investigators were also examining a separate case involving another 17-year-old following an FBI alert about an alleged plan to kill someone who had insulted his father.

Asked about the legal challenges facing the NCCIA, the official, who requested anonymity, said the father of the accused teenager in Bahawalpur had forgiven his son, leaving investigators with limited options to pursue the case.

He also said AI companies were under no legal obligation to alert Pakistani authorities, including the Federal Investigation Agency (FIA) and NCCIA, about suspected criminal activity on their platforms.

EXISTING LEGISLATION

Independent experts are divided over the need for AI-specific legislation.

Usama Khilji, director of digital rights organization Bolo Bhi, said existing criminal laws already covered offenses such as conspiracy and attempted murder, making separate AI-specific legislation unnecessary in many cases.

“Conspiracy and intention to murder is already a crime and AI is only a tool being used to get information about getting the crime,” he told Arab News. “There is no need for separate legislation.”

He said AI companies also needed to strengthen safeguards against criminal misuse of their platforms, adding that Pakistani authorities could engage with foreign companies even where local legislation had limited reach over firms based overseas.

Minahil Ali, an advocate of the High Courts specializing in artificial intelligence, legal technology and regulatory technology, said Pakistan was falling far behind in developing enforceable safeguards for AI.

“Pakistan remains drastically behind in establishing enforceable guardrails for artificial intelligence, struggling to move beyond basic comprehension while global technology has already crossed into the era of Agentic AI,” she told Arab News.

She said advanced AI systems were creating new challenges for criminal investigations.

“While local institutions are still coming to grips with simple text-generation risks, these autonomous capabilities blur traditional boundaries of criminal intent, bypass standard safety filters, and create severe attribution gaps that Pakistan’s regulatory framework is completely unequipped to address,” she said.

Ali warned that Pakistan’s tendency to introduce laws only after a crisis had emerged could leave authorities unprepared for more sophisticated AI-enabled threats.

“To avoid catastrophic failures, the state must urgently close structural gaps in its criminal justice system before autonomous threats materialize,” she said.

She said a key priority should be to modernize the legal framework governing attempted crimes, including Sections 324 and 511 of the Pakistan Penal Code and relevant provisions of the Prevention of Electronic Crimes Act (PECA), to clarify how criminal intent can be established in cases involving AI.

Sections 324 and 511 of the Pakistan Penal Code deal with criminal attempts, specifically attempted murder and attempts to commit other offenses, respectively.

Ali said investigators must distinguish between legitimate digital activity, such as asking an AI system questions for academic or creative purposes, and actions that demonstrate a genuine attempt to commit a crime.

“Proving intent requires mandatory SOPs [standard operating procedures] that mandate corroborating real-world evidence, such as material procurement or deployed API [application programming interface] scripts, rather than relying on isolated chat logs or false-positive safety flags,” she said.

Ali also called for updating the Law of Evidence, training specialized cyber-forensics units within the NCCIA to examine AI prompt histories and attempts to bypass safeguards, and establishing structured cross-border data-sharing arrangements with foreign AI companies.