Back A Shadow Intelligence Save as PDF
Chapters (58)
  1. Cover
  2. Also by Oliver Harris
  3. ONE
  4. TWO
  5. THREE
  6. FOUR
  7. FIVE
  8. SIX
  9. SEVEN
  10. EIGHT
  11. NINE
  12. TEN
  13. ELEVEN
  14. TWELVE
  15. THIRTEEN
  16. FOURTEEN
  17. FIFTEEN
  18. SIXTEEN
  19. SEVENTEEN
  20. EIGHTEEN
  21. NINETEEN
  22. TWENTY
  23. TWENTY-ONE
  24. TWENTY-TWO
  25. TWENTY-THREE
  26. TWENTY-FOUR
  27. TWENTY-FIVE
  28. TWENTY-SIX
  29. TWENTY-SEVEN
  30. TWENTY-EIGHT
  31. TWENTY-NINE
  32. THIRTY
  33. THIRTY-ONE
  34. THIRTY-TWO
  35. THIRTY-THREE
  36. THIRTY-FOUR
  37. THIRTY-FIVE
  38. THIRTY-SIX
  39. THIRTY-SEVEN
  40. THIRTY-EIGHT
  41. THIRTY-NINE
  42. FORTY
  43. FORTY-ONE
  44. FORTY-TWO
  45. FORTY-THREE
  46. FORTY-FOUR
  47. FORTY-FIVE
  48. FORTY-SIX
  49. FORTY-SEVEN
  50. FORTY-EIGHT
  51. FORTY-NINE
  52. FIFTY
  53. FIFTY-ONE
  54. FIFTY-TWO
  55. FIFTY-THREE
  56. FIFTY-FOUR
  57. FIFTY-FIVE
  58. ACKNOWLEDGMENTS

And no sign of Joanna.

I watched the road beneath the hotel window, then the park across the road, then drew the curtains again. I felt the clean coldness that came when I was about to act.

I opened my laptop – logged in to my agent’s Deutsche Bank account.

Balance $1,792,021.

I’d watched these figures a lot over the last year: debits and transfers signifying arms and alliances. I’d never stuck my hand in before, but it was simple enough. I clicked ‘transfer money’, went into payees and found the account details for my own Shariah-compliant account with the Jersey branch of the Abu Dhabi Commercial Bank. I set up three transfers of ninety grand, clicked ‘send now’.

It cleared in minutes.

I transferred a hundred thousand from Jersey to an account with Union Bancaire Privée in Geneva, then converted the rest of the money to Bitcoin, stored in an anonymous Dark Wallet account. I sent 50k to Evotec’s Cyprus account and submitted two tasking requests: information on the TutaNota address that had sent the message – where it had been set up and what device the email was sent from. I also sent the video file and asked for a forensic analysis: what kind of device was meant to have produced it; what CGI software was used in its construction.

I did my own research. The software involved used deep-level machine learning: you fed in images of someone’s face and the program taught itself how that face worked, how to generate new versions. It split itself into creator and critic, creating examples and then seeing if it could convince itself that they were real, feeding back errors until it could.

You needed the bodies, though, onto which the faces were mapped. Who was my body double? More seriously: who had the resources for all this? The process required a high-speed graphics processing unit. Supplying it with adequate training data meant having access to hundreds of images of that person. For someone moving around, like I was in the clip, seen from all angles, you might need thousands to be convincing.

I tried to sleep, got up an hour later and watched the clip again. I took a screengrab of my face and used it to set up a recognition alert on Cognizance, the only facial-recognition software that tracked online appearances rather than just CCTV feeds. I did the same for my companion. My sleep-deprived brain was convinced: this was Catalyst, the man who was going to kill me.

SIX