CIA Plans To Replace Spies With AI (thenextweb.com)

Human spies could soon be relics of the past. Dawn Meyerriecks, CIA's deputy director for technology development, recently told an audience at an intelligence conference in Florida that CIA was adapting to a new landscape where its primary adversary is a machine, not a foreign agent. From a report: Meyerriecks, speaking to CNN after the conference, said other countries have relied on AI to track enemy agents for years. She went on to explain the difficulties encountered by current CIA spies trying to live under an assumed identity in the era of digital tracking and social media, indicating the modern world is becoming an inhospitable environment to human spies. But the CIA isn't about to give up. America's oldest spy agency is transforming from the kind of outfit that sends people around the globe to gather information, to the type that uses computers to accomplish the same task more efficiently. This transition from humans to computers is something the CIA has spent more than 30 years preparing for.

Comments

Why not go the whole hog ...

[Alain Williams]

produce a VR world where the spies can chase after each other and leave the rest of us alone.

What could possibly go wrong?

[K. S. Kyosuke]

I thought CIA had already been deficient enough when it came to HUMINT. Now they propose to scale even further back when it comes to people immersed in the environment and attuned to the meaningful signals and things out of order? I can't see what could possibly go wrong with that...

Re:What could possibly go wrong?

[physicsphairy]

The CIA being the CIA, it's a good policy to let other countries think that they are dispensing with their human spies if, in fact, they aren't (so they receive less scrutiny). It's also an advantage to paralyze enemy's use of technology by making them paranoid. Or to cause them to give away the location of sensitive data by clumsily trying to protect it against a new threat. And if the enemy doesn't respond that way, then the CIA can continue development and reap the actual benefits of wielding AI against an unprepared enemy. I wouldn't take information they publicly disclose too literally.

Re:What could possibly go wrong?

[AHuxley]

The problem for the CIA is finding people who where never on social media and who have the needed international skills.
Every useful graduate that has language, arts, science to a standard that can pass as educated internationally had some online history.
What nations that the CIA is spy on did was buy into a lot of US social media data and build a large real time database of decades of US education.
Anyone in the USA who is educated to any useful "CIA" standard is now online as part of the US education system.

Great for the US education system and the graduates looking for work.
No so good for the CIA trying to pass a CIA spy as part of the US diplomatic workforce for decades of embassy work.
That persons past does not fit with the history presented and the work done in the embassy.
Other nations the CIA likes to spy on have created and contracted a lot of US detective agencies and US investigators who do pre-employment screening on workers. All US workers.
Their real time decades of data sets allow most nations to build up a database of every US graduate and "worker" with the academic ability to work internationally.

The CIA attempted to create past factual histories online for missions but if the data was not in place years ago and in the hands of the private investigators it does hold up. New data given to US social media stands out when placed in trying to look a decade old.
The only way around this is to use real US citizens who get asked to work for the CIA in another nation.
Business, NGO, charity, people of faith, engineers, historians, tourists, medical experts all get approached to be "CIA" and report back on their visit to another nation the CIA is interested in. Their past is perfect as it is all "true" and other nations are trusting of average people.

The AI is just an internal CIA database of all the normal people from everyday US who are now working for the CIA around the world.
Its not real spies the CIA knows and trusts so its "artificial intelligence" as in the people doing the spying for the USA are not classic CIA spies in any way.

A CIA computer system gets asked to find the best US graduates to spy on China. That group of people get approached and the CIA offers them a deal before their next trip to China.
That person works, holidays, studies in China with a few side tasks for the CIA and then returns to the USA with their reports.

The problem for the CIA is that it cant trust the people it is now contracting to be spies.
They did not come from within the CIA and all its own investigations.
Split loyalty and Communist support then becomes a risk for the CIA when approaching US academics to spy for the CIA.

Social media and its images upset decades of easy CIA spying globally.

They never learn...

[hyades1]

Go back half a century or so, and substitute "satellite" for "AI". The CIA let its efforts to recruit and/or train spies suffer to invest in various means of electronic information gathering. Training and recruiting is difficult, expensive and requires high-level decisions about the reliability of an agent and/or his information. This leads to a level of accountability many career civil servants don't necessarily want.

Unfortunately the change of direction meant the CIA lost its place as one of the top intelligence agencies in the Free World. I'm not saying it was horrible or incompetent, just not necessarily the place to go for a comprehensive picture of what was happening inside Russia, China and countless ugly little dictatorships.

So now so-called "artificial intelligence" is going to save the day?

It won't. The single best way to get intelligence is to have a lot of well-concealed, top level spies who have access to it. I doubt that will change for a long, long time.