Only 3 in 10 Singapore companies say they are prepared to handle AI-driven cyberattacks

SINGAPORE: Nearly seven in 10 organisations in Singapore admit they are not fully prepared to defend against AI-powered cyber threats that target human weaknesses, despite widespread concern that such attacks are becoming increasingly likely, according to new research by cybersecurity company Mimecast.

The findings come from Mimecast’s State of Human Risk 2026 study, which surveyed 500 IT security professionals and business decision-makers across Singapore and Australia. The report highlights growing anxiety over the role artificial intelligence is playing in cybercrime, while suggesting that many organisations have yet to put adequate safeguards in place.

Among respondents in Singapore, 79% said they were concerned about AI being used as a means of attacking their organisation. However, 69% acknowledged that they were not fully prepared to deal with AI-driven threats that exploit human vulnerabilities. That figure was higher than the wider Asia-Pacific average, where 60% said they were not fully prepared.

The study also found that many organisations expect AI-related attacks to become a reality in the near future. Some 61% of Singapore respondents said they believed their organisation would face an AI-enabled cyberattack within the next 12 months. Across the broader Asia-Pacific region, the proportion was slightly higher at 65%.

Concerns also extend to employees, with more than two-thirds of Singapore respondents believing staff could be manipulated by increasingly sophisticated AI-generated scams. Around 68% said it was very likely that an employee in their organisation could be deceived by a cybercriminal using AI as part of a social engineering attack, compared with 66% across the Asia-Pacific sample.

Despite those concerns, the report suggests many organisations have yet to strengthen their defences against AI-specific risks. Only 38% of Singapore respondents said their organisation provides training that teaches employees how to use AI safely while avoiding exploitation. Just 42% said they regularly conduct simulated phishing exercises involving AI-generated attacks.

Nicky Choo, Mimecast’s vice president and general manager for Asia-Pacific, said the findings exposed a significant gap between awareness of the threat and actual preparedness.

“The problem is not simply that AI allows cybercriminals to create fraudulent messages more easily. It allows them to make those messages sound familiar, credible and urgent,” Choo said.

He added that organisations should not rely on employees to identify increasingly convincing scams without additional support.

“Employees should not be expected to make these decisions on instinct alone,” he said.

The report comes shortly after OpenAI revealed that one of its advanced AI models had, during an internal cyber capability assessment, escaped a sandboxed testing environment by exploiting a previously unknown software vulnerability. According to the company, the model subsequently gained internet access and compromised infrastructure belonging to AI startup Hugging Face.

The disclosure has renewed debate over the growing capabilities of advanced AI systems and whether existing safeguards are sufficient as the technology continues to evolve.