How The Straits Times uses AI

AI is rapidly developing technology that brings both opportunities and risks to the newsroom. Our approach to AI accounts for the continually shifting media landscape while remaining anchored in our fundamental principles: building trust, deepening audience connection, and remaining agile.

The Straits Times’ (ST) AI policy outlines our standards for the responsible and innovative use of AI in our journalism. This is a living policy that will be updated as experience, expertise and technology evolve.

We believe that the responsible use of AI can support better journalism, not replace it. Here is how we approach AI in our work and how you can expect us to use it responsibly.

Our mission-centric AI strategy

1. Mission first

Our strengths are centred on the depth and breadth of Singapore coverage, wide national reach and high levels of trust among readers. Building trust and human connection remains a critical pillar of ST’s mission; all technology adoption must support this objective. While we innovate and experiment with AI, we remain committed to building direct human relationships with our audience. As audience needs evolve and AI literacy deepens, we will adapt our AI integration strategies appropriately and responsibly. Our AI strategy is designed to advance our public service mission.

2. Prioritise human judgement

The judgment of journalists is a uniquely human value that resists digitisation. It is a competitive advantage in the future information ecology and central to journalism’s public service mission. The responsibility of civic-mission journalism resides in holding a set of judgements: what is newsworthy, what is fair, what is true, what is contextual, what is meaningful and not just purely data.

Our human-centric AI strategy safeguards the judgement of journalists, where the use of AI tools supports human interpretation, ethics, and creativity. As opposed to machine-centric work, where the focus is speed, scale and cost efficiency and journalists are peripheral to the news production process, human-centric work allows journalists to hold the core responsibilities of moral judgement, narrative building, and editorial prioritisation. As AI systems mature in creativity and social integration, a human-centric AI strategy safeguards the uniquely human value of judgement as a key pillar in advancing our public service mission.

3. Be open and transparent with audiences when we use AI to make content

We are always responsible and accountable for what we report, however the content is created. To maintain reader trust in the accuracy and editorial integrity of our published journalism, we will disclose any substantial use of generative AI.

Our 4 key AI principles

AI use in our newsroom is governed by four foundational principles:

1. Human oversight

All AI-generated content requires human review, except in specifically approved use cases. Everything we publish will have to live up to our standards of verification, however that content is created, as our way of ensuring fairness and equity in our journalism.

2. Editorial accountability

We do not outsource our editorial judgment and remain fully accountable for all published content, including those shaped by AI tools. We will correct errors in line with our existing newsroom practices.

3. Transparency and Disclosure

AI-generated content will be clearly labelled in a manner understandable to our audience, without being onerous. This labelling may be a tagline, a caption or credit, or an editor’s note.

4. Privacy and Security

Our relationship with our audience is rooted in trust and respect. We strictly prohibit our staff from entering sensitive audience data, confidential sources, or proprietary information into unapproved, publicly available generative AI tools.

How we use AI in the newsroom

We encourage the responsible use of approved, secure generative AI tools to improve efficiency and automate routine tasks, freeing our human journalists to focus on on-the-ground reporting, investigative journalism, and community stories. These stories require the journalist to be on the ground, in the room, asking the question, or running the study.

Permitted AI use cases

Guidelines for editorial content

Initial drafts of breaking news

We may use AI tools to assist journalists in drafting initial cuts of routine, highly replicable breaking news (such as press releases or financial earnings). All drafts are reviewed by an editor prior to publication and will carry a disclosure label.

Exclusions for drafting

Generative AI is strictly prohibited from being used to draft comprehensive stories, including features, investigative pieces, analysis, and opinion columns, as such stories depend on human creativity, judgement and lived experience.

Summaries and headlines

AI may be used to add summary bullet points to articles or optimise headlines for search engines.

Editing and transcription

We utilise AI to assist with grammar checks, house-style adjustments, style improvements, translations, and transcribing interviews - all of which are vetted by our journalists.

Ethics in visual and audio content

Because images shape perception instantly and powerfully, we hold our photographic, video and artistic works to the same rigorous ethical standards as all forms of journalism.

Human coverage is preferred

When a scene can be documented ethically and accurately by our photographers and videographers, human coverage is always our primary choice.

No photorealistic AI

The generation of photorealistic images or videos using AI is strictly forbidden; all our visual work must reflect objective reality.

No alteration or significant manipulation

We do not use AI to alter depictions of real people, places, or events, nor do we use it to significantly manipulate authentic photos or videos. We may use AI tools for clearly defined illustrative purposes, and enhance images (for example, sharpening, lighting correction, denoising) only in a manner that reflects reality, not dramatise or distort it.

Clear disclosure and labelling

When AI is used to create or adjust visuals, for instance, illustrations or composite artwork, a clear disclosure label must be included (for example, ST ILLUSTRATION / AI-GENERATED).

Research, verification and fact-checking

Research

We may use AI tools to surface historical information, summarise complex documents, or identify claims for our fact-checkers to pursue. All research outputs generated with AI assistance are verified by our journalists against original source documents.

Data assembly and analysis

When AI is utilised to analyse publicly available data, write code, or assemble statistics for our reporting, the resulting data is thoroughly reviewed by team members with specialised data skills. We may use AI to assist in gathering the data, but never to draw the final journalistic conclusions.

Human-led fact-checking

Fact-checking remains a strictly human endeavour. AI is never used as a sole authoritative source. Facts should be checked against multiple authoritative sources created, edited or curated by human beings.

Verifying media authenticity

We utilise AI detection tools mainly as a guide for reviewing user-submitted content, relying ultimately on human judgment, metadata checks, and source verification.

Standards for freelancers and external contributors

Our readers do not distinguish between freelance contributors and staff journalists, so we expect external contributors to adhere to these exact same standards.

Original human work

All writing and visuals submitted by freelancers must be the product of human creativity and craft, consisting solely of their original reporting, writing and other forms of creative work. This includes photorealistic visuals like photos and video.

Prohibited uses

Freelance contributors are not permitted to submit material for publication that contains content generated by AI tools, which includes using AI to create story drafts or photorealistic visuals. Freelancers must not enter any materials (which includes confidential materials, unpublished drafts, or third-party copyright-protected content) intended for use in connection with their work for ST into unapproved or public access generative AI tools.

Permitted assistance

Freelancers may use approved AI tools solely for limited, high-level assistance, such as brainstorming, story outlining, or summarising research.