TOKYO -- Japan will seek ways to obtain information on undisclosed software vulnerabilities from U.S. artificial intelligence developers and other overseas tech companies at an early stage, as Tokyo scrambles to respond to a spike in cyberattacks, Nikkei has learned.

The government will present an emergency package at a ministerial cybersecurity meeting to be convened as early as Friday. Prime Minister Sanae Takaichi had directed officials to come up with measures at a meeting in late October.

After a vulnerability was discovered in the PHP programming language last month, many Japanese companies shut down their websites. In the U.S., most businesses kept their sites operating thanks to cooperation between government agencies and tech companies that possessed information about the vulnerability.

Japan's government, some companies and financial institutions already have access to advanced AI models like Anthropic's Claude Mythos. The government recently asked major U.S. AI developers for expanded access to their models in response to a recent wave of cyberattacks believed to have been enabled by AI.

But vulnerabilities known by U.S. companies are not necessarily being shared with Japan's government or private sector, Japanese officials said. If authorities could learn about such risks before they are disclosed publicly, cybersecurity responses could become faster.

Once serious vulnerabilities are revealed, attackers rush to exploit them. Early access would let Tokyo assess more rapidly how software flaws could affect government agencies and critical infrastructure such as electricity grids and telecommunications networks.

Companies can be reluctant to share cyberattack information with each other because revealing details may expose weaknesses in their systems.

To address such concerns, Tokyo plans to build a broad information-sharing framework, collecting intelligence not only from victimized companies but also from cybersecurity vendors, cloud service providers and other organizations on the front lines of cyberdefense.

The information would be centralized at Japan's National Cybersecurity Office, enabling authorities to grasp the overall scope of attacks across industries and regions and assess the risk of spillover to other companies. Tokyo aims to have the public and private sectors work together on analysis and responses to limit damage.

The emergency package also will include measures to encourage cybersecurity vendors and cloud service providers to join public-private coordination councils.

Those councils so far have focused on critical infrastructure operators. Bringing in security and cloud companies whose systems could spread the effects of attacks to other businesses is expected to help prevent damage from cascading.

Japan also is considering cybersecurity measures such as issuing guidance on the storage of images and other data collected by companies for identity verification, and working with the U.K.'s AI Security Institute to assess the cyberattack capabilities of advanced AI models.