Skip to content
Lacambra Web Portal Services
← 2026, week by week
Government · Week 3, 12 to 18 January 2026

Grok ban in the Philippines: deepfakes of your staff

The Philippines blocked Grok over sexual deepfakes and lifted the block five days later. The deepfake an office should worry about is of its own people.

A plain white porcelain mask with no features painted on it, lying on a dark desk beside a small hand mirror turned face down on a folded oxblood cloth.

On 15 January the Secretary of Information and Communications Technology, Henry Aguda, said the government was moving to block xAI's Grok chatbot over its ability to generate sexualised images, including content that could put children at risk. By the next day the National Telecommunications Commission, working with the Cybercrime Investigation and Coordinating Center (CICC), had ordered internet providers to block it, and the CICC cited the tool's ability to make deepfakes of real people without their consent. The block was lifted on 21 January, after xAI committed to new safeguards. For an office, the lasting lesson is less about one chatbot than about how easily a real person can be faked, and how quickly a tool can disappear.

What was blocked, and why it came back

The Philippines moved days after Indonesia, which had blocked Grok temporarily on 10 January. The CICC said its action was in line with Republic Act 10175, the Cybercrime Prevention Act, GMA News reported. The block lasted less than a week. According to Philstar, xAI pledged to change the tool for the local market, dropping the image manipulation features that allowed deepfakes and excluding pornographic content. The DICT said the company had also tightened restrictions on prompts involving real people, and the CICC said it would keep watching the app.

The fake an office should worry about

Grok's images made the news, but the warning that matters to an office came from the CICC on 21 January, the day the block ended. It told small and medium businesses that criminals were using AI to clone the voices and faces of business owners, to get employees and clients to transfer money or hand over sensitive information. For many organisations the raw material is already public: officers' photographs on the website, their speeches on video.

Training staff to spot a fake is a weak defence, because the better the fake, the less there is to spot. The stronger rule is that any request for money, passwords or records is confirmed through a channel the requester did not choose, such as a call back on a number already on file. A cloned voice works best where urgent requests already skip the official route, so that shortcut needs the same rule.

What it means for approvals and AI tools

Approval systems help here, because they take the decision out of the phone call. The approval workflows I build take a request past each person who has to sign it off, with a record of who signed off what. A cloned voice can put pressure on someone, but the approval still has to be given in the system, on the record, by the person whose stage it is.

The block also showed that an AI tool can vanish within a day. A short written policy covers both risks: which AI tools staff may use for work, what may be pasted into them, and a plain rule that nobody uses AI to make images, video or audio of real colleagues, members or clients. A colleague's photograph is information from which their identity is apparent, which is how the Data Privacy Act of 2012 (Republic Act 10173) defines personal information, so that rule is a privacy rule too.

Five rules worth writing down

  • Name the AI tools staff may use for work, and keep the list short enough to remember.
  • Forbid using AI to make images, video or audio of real colleagues, members or clients.
  • Confirm every request for money, passwords or records on a channel the requester did not choose, however senior the voice.
  • Make sure no process depends on one consumer chatbot, and know what you would do if it were switched off overnight.
  • Tell staff where to report a suspected deepfake: the CICC takes reports on its 1326 hotline.

If your approvals still happen in phone calls and chat messages, the fit check will tell you whether a proper approval system is worth building.

Sources

  1. Reuters, via The Star, Philippines seeks to block access to Grok on child safety concerns, 15 January 2026
  2. GMA News, PH blocks Grok, xAI proposes corrective measures, 16 January 2026
  3. DZRH News, CICC warns SMEs of rising deepfake cloning scams, 21 January 2026
  4. The Philippine Star, CICC to lift ban on Grok AI, 22 January 2026
  5. Newsbytes.PH, DICT lifts ban on Grok after safeguards, compliance commitments, 23 January 2026
  6. Lawphil, Republic Act No. 10173

Enrique Lacambra III builds internal business systems from Cainta, Rizal.

Email me← All writing

Tell me what is broken and I will tell you if it is worth building. A 20-minute call, no charge, and I will say so if you do not need me.