{
  "id": 11438658,
  "title": "Existing safeguards by AI companies insufficient as models grow more capable: Josephine Teo",
  "url": "https://urgent.news/2026/10/02/existing-safeguards-by-ai-companies-insufficient-as-models-grow-more-11438658",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-02T12:25:49.000Z",
  "source": {
    "name": "Channel News Asia",
    "slug": "channel-news-asia",
    "url": "https://www.channelnewsasia.com/singapore/ai-safeguards-insufficient-content-filters-josephine-teo-6426716"
  },
  "original_language": "en",
  "account": "Minister Teo stresses AI safeguards are not enough as models grow\n\nSingapore's digital development and information minister Josephine Teo warned on Friday that existing safeguards put in place by AI companies are insufficient to address the risks posed by increasingly capable AI systems. In a Facebook post, Teo said malicious users could misuse powerful AI to cause harm, such as by finding vulnerabilities in computer systems, writing malicious code, automating parts of cyberattacks and making scams more convincing.\n\nSafeguards like content filtering, limiting AI model access, pre-release testing, user verification, misuse detection and account suspensions are important, but not enough, Teo said. Malicious users can still employ openly available models that anyone can download and run independently, making it easier to bypass safeguards and hide misuse. The minister emphasized the need for multiple layers of defense, including stronger cyber defenses and a better understanding of advanced AI systems' capabilities and limitations.\n\nTeo highlighted the importance of stronger cyber defenses, better detection and recovery capabilities, especially for government systems and essential services. She also stressed the importance of balancing convenience and security, noting that the Cyber Security Agency of Singapore has issued guidance on patching vulnerabilities, using strong authentication and tightening access controls.\n\nOrganizations can use AI defensively to identify vulnerabilities before attackers do, Teo said. Even without the most advanced AI models, organizations can leverage available AI tools to improve their cyber defense. The minister emphasized that safeguards become crucial when organizations give AI models access to their data, tools, and processes, as AI agents can misunderstand instructions, pursue unintended goals, be tricked by malicious information, or be given excessive authority.\n\nSingapore needs multiple lines of defense, including stronger cyber defenses, clear limits on AI capabilities, appropriate human oversight, and the ability to test, monitor, and learn as the technology evolves. Teo also called for Singapore to go \"further upstream\" to better understand the behavior of increasingly capable AI models and their limitations. The country's AI Safety Institute is building technical capabilities to evaluate advanced AI systems, and Singapore is interested in collaborating with international experts to develop stronger safeguards and technical standards.",
  "summary": "Content filters, user verification and pre-release testing are among existing safeguards, but malicious users can bypass them by turning to openly available AI models, says the digital development and information minister.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "CNA - Singapore",
        "title": "Existing safeguards by AI companies insufficient as models grow more capable: Josephine Teo",
        "url": "https://urgent.news/2026/10/02/existing-safeguards-by-ai-companies-insufficient-as-models-grow-more",
        "published": "2026-10-02T12:25:49.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}