Latest / AI Newspaper / UK launches Information Defence centre and pitches global frontier-AI standards at the UN

Transcript

UK launches Information Defence centre and pitches global frontier-AI standards at the UN · AI Newspaper

• 0:00
  Welcome to the AI Newspaper, your daily update
• 0:02
  on everything happening with AI. Visit AINewspaper
• 0:06
  .org to see today's stories and sign up for email
• 0:09
  alerts. Always great to be here. So I want you
• 0:11
  to imagine waking up, you grab your phone, and
• 0:14
  you see this breaking news alert from the BBC.
• 0:17
  Just a standard morning routine. Exactly. And
• 0:19
  the logo is perfect. The font is identical. And
• 0:22
  the article is describing a catastrophic collapse
• 0:24
  in the local banking sector. Oh, wow. Yeah. So
• 0:27
  obviously you panic, right? You share it in group
• 0:29
  chat. You text your family. But the article isn't
• 0:32
  real at all. The BBC never published it. Right.
• 0:35
  It was entirely fabricated by an automated system
• 0:37
  deployed by a foreign intelligence agency. And
• 0:40
  it was designed specifically to target your demographic
• 0:42
  because some algorithm identified you as, you
• 0:45
  know, economically. Which sounds completely dystopian.
• 0:49
  It does, but that is not a sci -fi pitch. That
• 0:51
  is the exact scenario that UK Prime Minister
• 0:54
  Andy Burnham just laid out at the UN General
• 0:57
  Assembly. Yeah, he basically drew a line in the
• 0:59
  sand in New York. We are looking at this head
• 1:03
  -on collision between the rapid advancement of
• 1:05
  frontier AI models and the remarkably slow gears
• 1:09
  of international diplomacy. Right. Burnham is
• 1:11
  essentially casting AI as this dangerous multiplier
• 1:15
  of state -level information warfare. But at the
• 1:18
  exact same time, he is pitching Britain as the
• 1:21
  global honest broker. You know, the one capable
• 1:23
  of establishing global safety standards. Which
• 1:26
  is a really fascinating tightrope to walk. Yeah.
• 1:28
  Especially when you look at the glaring structural
• 1:30
  tension underlying the entire industry right
• 1:33
  now. Right. Because the major AI labs building
• 1:36
  these systems currently operate largely outside
• 1:38
  of mandatory government scrutiny. Exactly. So
• 1:41
  to understand why the UK is pushing so hard for
• 1:43
  global AI standards, we first have to look at
• 1:46
  the immediate threat Burnham highlighted at the
• 1:48
  UN. The weaponization of AI by hostile states.
• 1:51
  Yeah. He announced the creation of a new agency,
• 1:53
  the National Center for Information Defense.
• 1:56
  And its mandate is highly specific. It brings
• 1:59
  together intelligence agencies, law enforcement,
• 2:01
  social media platforms, all to detect, attribute,
• 2:04
  and disrupt AI -enabled disinformation campaigns
• 2:08
  and deepfakes. And he didn't hold back or speak
• 2:11
  in hypotheticals about who is running these campaigns
• 2:13
  either? No, he did not. He specifically called
• 2:15
  out Russia. He stated the Kremlin spends around
• 2:18
  1 .3 billion pounds annually on... information
• 2:21
  manipulation. Which is a staggering number. I
• 2:24
  mean, when you break down what that kind of budget
• 2:26
  buys in the age of generative AI, it's just massive.
• 2:31
  It really is. Burnham listed the tactics right.
• 2:33
  Using bot networks, standing up completely fake
• 2:37
  websites, and forging the branding of 28 different
• 2:40
  British organizations. Including universities,
• 2:42
  and like your example earlier, the BBC. Right.
• 2:45
  He even pointed to evidence of attempted interference
• 2:47
  going all the way back to the 2019 UK general
• 2:50
  election. Though we should definitely note that
• 2:51
  Russian officials have completely rejected all
• 2:53
  of these allegations. Yeah, they push back hard.
• 2:56
  Foreign Minister Sergei Lavrov characterized
• 2:59
  the Western media claims as anti -Russian cliches
• 3:03
  and fantasies. He basically dismissed the whole
• 3:06
  thing as Russophobia. So you have this really
• 3:08
  intense diplomatic clash playing out over what
• 3:11
  is objectively true. But looking at the mechanics
• 3:14
  of the UK's response, I kind of look at this
• 3:16
  new national center as almost like A national
• 3:20
  digital immune system. Oh, that's a good way
• 3:22
  to frame it. Yeah, you have intelligence agencies
• 3:24
  trying to fight off this rapidly mutating virus
• 3:27
  of AI fakes. Because traditional cybersecurity
• 3:30
  is usually about stopping a hacker from breaching
• 3:33
  a server to steal data, right? Right, patching
• 3:35
  firewalls and that sort of thing. Exactly, but
• 3:37
  this is completely different. How do intelligence
• 3:39
  agencies actually disrupt an AI -generated narrative
• 3:42
  that is already out in the wild, already being
• 3:45
  shared by regular citizens? It requires a massive
• 3:47
  shift in tactics. I mean, if a forged article
• 3:50
  is completely indistinguishable from reality,
• 3:52
  scanning the text itself doesn't actually help
• 3:55
  you. No, the text looks perfect. So the defense
• 3:57
  mechanism has to shift toward identifying the
• 3:59
  distribution network. Intelligence agencies have
• 4:02
  to trace the subtle digital footprints of the
• 4:05
  bot farms, amplifying the post. Or they have
• 4:07
  to collaborate directly with social media companies
• 4:10
  to throttle the algorithmic reach of a deepfake
• 4:13
  before it goes viral. Exactly. It's an acknowledgment
• 4:16
  that the battlefield has fundamentally shifted
• 4:19
  from traditional military engagements to cognitive
• 4:22
  spaces. A distorted, artificially amplified narrative
• 4:26
  can paralyze a population just as effectively
• 4:29
  as a physical. attack on a power grid. And Burnham
• 4:32
  actually dug into why those narratives take root
• 4:34
  so effectively in certain populations, which
• 4:37
  I thought was a really interesting pivot. Yeah,
• 4:38
  he made this pointed connection between societal
• 4:41
  inequality and susceptibility to disinformation.
• 4:44
  Right. He argued that people struggling with
• 4:46
  the cost of living or, you know, young people
• 4:48
  not in education, employment or training, they
• 4:52
  are particularly vulnerable to these narratives
• 4:54
  of decline. Burnham is diagnosing a fundamental
• 4:57
  societal fragility there. Because an AI -generated
• 5:01
  deepfake doesn't work if the audience completely
• 5:03
  rejects the underlying premise. Yeah, if things
• 5:05
  are going great, you don't believe the sky is
• 5:07
  falling. Exactly. But if you have citizens who
• 5:10
  are struggling to afford basic groceries or who
• 5:13
  feel locked out of the housing market, they are
• 5:15
  primed to believe a narrative that says the system
• 5:18
  is collapsing. Or that a specific institution
• 5:21
  is corrupt. So the algorithm identifies that
• 5:25
  you, the listener, are pausing to read articles
• 5:27
  about inflation or housing costs. And then it
• 5:30
  feeds you a fabricated story specifically engineered
• 5:33
  to exploit your existing financial anxiety. The
• 5:36
  technology is really just the accelerant being
• 5:39
  poured on a pre -existing fire, which leads to
• 5:41
  my question. Is Burnham treating AI disinformation
• 5:44
  purely as a tech problem or is he actually diagnosing
• 5:48
  a deeper societal issue? I think he's definitely
• 5:50
  pointing to the societal issue. He is basically
• 5:52
  arguing that building a national immune system
• 5:55
  requires a lot more than just catching the deep
• 5:58
  fakes. It requires actual economic reform. Right,
• 6:00
  creating actual economic opportunities so people
• 6:02
  aren't susceptible to a toxic automated narrative
• 6:05
  in the first place. AI disinformation is just
• 6:08
  way more effective when it finds a receptive,
• 6:10
  vulnerable host. So with AI supercharging these
• 6:13
  global information wars, Burnham used his UN
• 6:16
  speech to position the UK as the geopolitical
• 6:19
  bridge capable of establishing global safety
• 6:22
  rules. Which we really need to contextualize
• 6:25
  this pitch. Yeah, let's talk about that. the
• 6:26
  timing. This was Burnham's very first multinational
• 6:29
  meeting since taking power. And to put the UK's
• 6:32
  recent political turbulence in perspective, he
• 6:35
  is the country's seventh prime minister in a
• 6:37
  single decade. Seventh in 10 years. That is a
• 6:40
  lot of turnover. It is. So he used this massive
• 6:43
  platform in New York, not just to warn about
• 6:46
  AI, but to actively try and rebuild Britain's
• 6:49
  diplomatic status, you know, projecting stability
• 6:51
  on the world stage. He pitched the UK as an honest
• 6:54
  broker for global AI standards, and he cited
• 6:57
  a few key geopolitical advantages to back this
• 7:00
  up. Right. Like Britain's historically close
• 7:02
  ties with the U .S., its improving relationship
• 7:04
  with the EU, and relatively decent relations
• 7:07
  with China. He even announced plans to put AI
• 7:11
  governance right at the heart of the UK's G20
• 7:13
  presidency next year. It's very strategic positioning.
• 7:16
  He is leaning heavily into the infrastructure
• 7:19
  the UK has already built up over the last couple
• 7:21
  of years. He really highlighted the UK's AI Security
• 7:25
  Institute. Which was actually established under
• 7:27
  his predecessor, right? Yeah, and he described
• 7:29
  it to the UN as world leading. And Burnham pointed
• 7:32
  out that the UK Institute already works closely
• 7:34
  with the US in leading AI labs on safety. Plus,
• 7:37
  he announced a new AI and autonomy partnership
• 7:40
  between the UK and the US. Focused on protecting
• 7:43
  critical national infrastructure. So things like
• 7:45
  power grids and water systems. OK, but I have
• 7:48
  to offer some pushback on this diplomatic framing
• 7:50
  because it sounds great on paper, right? Positioning
• 7:53
  the UK as the ultimate diplomatic middle child.
• 7:56
  It's a nice narrative. But can one country realistically
• 7:59
  act as a broker to rein in AI when superpowers
• 8:03
  like the US and China are locked in this intense
• 8:06
  race for dominance? It is an incredibly tough
• 8:09
  sell. AI development is currently viewed by the
• 8:12
  major powers entirely through the lens of a new
• 8:14
  arms race. Right. Whoever gets there first wins.
• 8:17
  Exactly. The economic and military advantages
• 8:20
  of achieving artificial general intelligence
• 8:22
  first, or even just dominating the current generation
• 8:24
  of frontier models, are just too vast. So slowing
• 8:29
  down for the sake of an international standard
• 8:31
  brokered by a third party, even an ally like
• 8:34
  the UK, that kind of runs counter to their immediate
• 8:37
  national security imperatives. It absolutely
• 8:40
  does, which brings us right to the stark structural
• 8:43
  reality of the AI industry right now. The massive
• 8:46
  gap between government ambition and mandatory
• 8:48
  scrutiny. Yes, because look at what Burnham actually
• 8:52
  demanded at the UN. He declared that fundamental
• 8:54
  decisions about AI must be taken by elected governments,
• 8:57
  not left to tech companies. He demanded full
• 9:00
  and total visibility of new frontier models.
• 9:03
  But the reality check here is that testing at
• 9:04
  the UK's much -touted AI Security Institute is
• 9:07
  currently entirely voluntary. Entirely voluntary.
• 9:10
  Meaning if major labs like Anthropic can simply
• 9:13
  say no thanks to voluntary safety tests, isn't
• 9:16
  the UK essentially trying to enforce a global
• 9:18
  speed limit without installing any traffic cameras?
• 9:21
  That is the perfect analogy. And the car manufacturers
• 9:23
  in this scenario have massive financial incentives
• 9:27
  to keep the hood completely shut. Because these
• 9:30
  frontier models cost billions of dollars to train.
• 9:32
  Exactly. The architecture, the training data,
• 9:36
  the specific mathematical weights, that is their
• 9:38
  core intellectual property. So letting a foreign
• 9:41
  government security institute poke around inside
• 9:43
  their most advanced system before release presents
• 9:46
  a huge perceived risk of corporate espionage.
• 9:49
  Or just leaks. And if a major lab doesn't want
• 9:52
  to show the UK institute what they're building,
• 9:54
  the UK has absolutely no legal mechanism to force
• 9:57
  them. None at all. And we are already seeing
• 10:00
  the limitations of this voluntary approach play
• 10:02
  out. It was reportedly just revealed that the
• 10:05
  major AI company Anthropic completely refused
• 10:08
  to submit its latest model for testing at the
• 10:10
  UK Institute. Which is huge. The implications
• 10:13
  of a leading lab just opting out highlight exactly
• 10:15
  how toothless voluntary safety agreements are.
• 10:17
  Right. When push comes to shove, if a company
• 10:20
  decides that a safety review will delay their
• 10:22
  product launch, they can just decline. And this
• 10:24
  realization is causing a lot of global friction.
• 10:27
  We just saw over 20 countries sign a Norway -led
• 10:31
  declaration calling for mandatory pre -release
• 10:34
  testing of frontier models. Countries like Germany
• 10:37
  and Canada signed on to that. But the US, China
• 10:39
  and the UK did not sign it. Which is putting
• 10:41
  Burnham under immense domestic political pressure.
• 10:44
  You have figures like former UK cabinet minister
• 10:47
  Darren Jones openly urging Burnham to sign it.
• 10:50
  Jones wants him to broker a mandatory test agreement
• 10:53
  with the U .S., right? Yeah. But showcasing the
• 10:56
  sheer difficulty of reaching that global consensus,
• 10:59
  we have to look at the U .S. perspective. Oh,
• 11:01
  this was wild. U .S. President Donald Trump used
• 11:03
  his U .N. press conference to explicitly reject
• 11:06
  any, quote, globalist scheme to control A .I.
• 11:09
  He called the warnings of A .I. destroying humanity
• 11:12
  a hoax, and he flatly stated that the U .S. will
• 11:14
  encourage it, not rein it in. It is a complete
• 11:17
  paradigm clash. The U .K. is pushing for international
• 11:19
  guardrails, treating AI as a shared global risk,
• 11:22
  while the U .S. administration is treating AI
• 11:25
  as a critical domestic asset that must be aggressively
• 11:28
  accelerated. And Trump also noted during that
• 11:30
  same U .N. appearance that previous U .K. government
• 11:34
  rhetoric had been too tough on Moscow. Which
• 11:36
  Downing Street firmly pushed back against. But
• 11:39
  it shows how strained the broader geopolitical
• 11:42
  alliances are right now. It's a terrifying contradiction
• 11:45
  when you step back and look at it. It really
• 11:47
  is. I mean, we'll leave you with this final paradox
• 11:49
  to mull over today. The UK is raising the alarm
• 11:52
  on the world stage that AI is a potent weapon
• 11:55
  for state level information warfare. Right. Actively
• 11:59
  exploiting the economic anxieties of everyday
• 12:01
  citizens. Yet the very nation's dominating AI
• 12:04
  development refused to commit to mandatory safety
• 12:07
  testing. So if governments won't legally compel
• 12:10
  tech labs to open their models for scrutiny,
• 12:12
  who Who is actually dictating the future of global
• 12:15
  security? Elected officials or the private companies
• 12:19
  building the technology? It's the million -dollar
• 12:21
  question. Thanks for listening to the AI Newspaper.
• 12:24
  Visit innewspaper .org to catch up on the latest
• 12:27
  updates and sign up for free email alerts.