The Theater of "AI Safety" vs. "AI Race": A Marxist-Leninist Analysis
Introduction: The Stage Is Set
In the previous article, The Debate on AI: A Marxist-Leninist Approach, we established the foundational
bourgeois narratives: AI as neutral, AI as utopia, AI as inevitable, AI
as something to be "regulated" within capitalism. We demonstrated
that the character of technology is determined not by its technical
specifications but by the relations of production under which it is
developed and deployed.
Since then, the ideological
battlefield has shifted. The dominant debate in bourgeois media is no longer
simply "is AI good or bad?" It is now framed as a choice
between two competing strategies:
- Slow down AI development to prioritize safety.
Industry researchers and executives warn that fast-moving AI could
"escape human control" or pose a "10% or greater
existential threat to humanity within a decade." Essays urge an
industry-wide slowdown in capability gains so that "safety guardrails
and oversight can keep pace." State attorneys general, mainstream
observers, and critics point to "concrete risks happening now,"
including "autonomous hacking incidents, deepfakes, job losses, and
surveillance expansion." They warn of "AI systems being used to amplify
or lead espionage, cyber-attacks or biological warfare, or disrupt
economies — whether by accident or design." They invoke "human
rights risks posed by AI."
- Race ahead unchecked to beat China. Others
argue that "decelerating AI hands a vital strategic advantage to
China." They advocate for "stronger cybersecurity and defensive
infrastructure over braking innovation." The logic is simple: if we
slow down, they speed up — and the first to achieve AGI will dominate the
21st century.
On the surface, these appear to be opposing positions. One calls for caution, the other for speed. One emphasizes safety, the other security. One fears existential risk, the other fears geopolitical defeat.
But this is not a debate.
It is a theater.
The script is written by the same
corporations that fund both camps. The stage is owned by the same
media that amplify both camps. The audience is the public, which is invited
to choose between two forms of control. The only question that is not
asked — who owns AI, and for what purpose? — is the only question that
matters.
This article will expose the
theater. It will show that the "safety" camp and the
"race" camp are not adversaries but accomplices — two moments of
the same bourgeois ideology, each performing a necessary function for capital.
The first creates the regulatory moats that entrench monopoly. The second
provides the geopolitical justification for state subsidies, export controls,
and technological hegemony.
The real danger is not AI.
The real danger is capitalist AI. And the real solution is not slower AI or
faster AI — it is socialist AI.
Two Camps, One Script
The "safety" camp
presents itself as the conscience of the tech industry. Its representatives —
former OpenAI employees, academic researchers, NGO directors, and sympathetic
journalists — warn that AI development is proceeding too fast for safety
measures to keep up. They invoke existential risk, autonomous hacking,
deepfakes and disinformation, job losses, surveillance expansion, biological
warfare, economic disruption, and human rights risks. These are presented as
universal concerns, threats to all humanity.
The "race" camp
presents itself as the realist alternative. Its representatives — venture
capitalists, defense intellectuals, national security officials, and
sympathetic journalists — argue that deceleration hands a vital strategic
advantage to China, that stronger cybersecurity and defensive infrastructure
are preferable to braking innovation, that AI is essential for national
security, and that the "AI safety" movement is naive. This camp does
not deny the risks. It simply argues that the risks of not racing are greater
than the risks of racing.
On the surface, the two camps
appear to be engaged in a genuine debate. The "safety" camp accuses
the "race" camp of recklessness. The "race" camp accuses
the "safety" camp of naivety. Each has its own experts, its own think
tanks, its own media platforms. Each claims to speak for humanity — one for its
survival, the other for its freedom.
But this is a performance. The
two camps are not adversaries. They are a division of labor. One creates
the regulatory moats; the other creates the geopolitical justification for
state subsidies and export controls. One says, "slow down so we can
control it"; the other says, "speed up so we can control it."
Both say control.
The theater of debate serves to
make control appear as a solution rather than as the problem.
What Both Camps Agree On
The two camps appear to disagree
on everything. But they agree on everything that matters.
Both accept that AI
development is driven by capital. Neither questions the ownership of the
means of production. Neither proposes public ownership of AI. Neither advocates
for democratic control of technology. Both accept that AI should be developed
by private corporations, for private profit.
Both accept that the purpose
of AI is profit. The safety camp wants "responsible" profit. The
race camp wants "competitive" profit. Neither wants to subordinate AI
to human needs. Neither asks: what would AI look like if it were developed to
serve the many, not the few?
Both accept that the working
class is passive. The safety camp wants to protect workers from AI. The
race camp wants to retrain workers for AI. Neither wants workers to control AI.
Neither asks: what would AI look like if it were developed by and for the
working class?
Both accept that the
nation-state is the primary unit of analysis. The safety camp wants
international regulation. The race camp wants national advantage. Neither wants
international solidarity of the working class. Neither asks: what would AI look
like if it were developed cooperatively by all nations, for the benefit of all
humanity?
Both accept that technology is
neutral. The safety camp wants to make AI "safe." The race camp
wants to make AI "fast." Neither wants to make AI
"socialist." Neither asks: what would AI look like if it were
subordinated to the relations of production of a socialist society?
Both conclude with
"control." The safety camp wants to control AI through
regulation, ethics committees, and "alignment." The race camp wants
to control AI through state subsidies, export controls, and geopolitical
competition. Neither wants to abolish control. Neither wants to democratize
control. Both want control — by the same corporations that already control
Western governments.
The debate is not between
"safety" and "race." It is between two forms of capitalist
control. The real alternative — socialist AI — is not on the stage at
all.
Who Writes the Script
The "safety" camp and
the "race" camp are funded by the same interests: tech monopolies,
which profit from both regulation and deregulation, as long as they remain
dominant; venture capital, which profits from rapid development and from
regulatory moats that eliminate competitors; the defense industry, which
profits from geopolitical competition and from AI-driven weapons systems; the
US government, which profits from technological hegemony and from the national
security state; and the media, which profits from advertising revenue from all
of the above.
These are not competing
interests. They are the same interests, pursuing different strategies. The
"safety" camp and the "race" camp are not adversaries. They
are accomplices.
The safety camp's real
function is regulatory capture. "Safety" regulations — licensing
requirements, audit trails, red-teaming mandates, liability insurance,
"alignment" certifications — are trivially affordable for a
trillion-dollar corporation. They are impossible for startups, academic labs,
open-source collectives, and public-interest projects. When a small team of
researchers releases an open-weight model, they cannot afford a compliance
department. When a university lab develops a new algorithm, they cannot afford
a legal team. When a cooperative builds a community-owned AI, they cannot
afford the insurance. The result: "Safety" regulations eliminate
competition. They entrench the monopolies. They are not a brake on capital —
they are its acceleration.
The "ethics committees"
and "responsible AI" frameworks that the safety camp proposes are
designed by and for the monopolies. Who sits on these committees? Former
executives of the monopolies. Who sets their agendas? The foundations funded by
the monopolies. Who defines "responsible AI"? The corporations that
profit from AI. "Responsible AI" is not a universal standard. It is
the standard of the ruling class, imposed on everyone else.
The "alignment"
discourse assumes that AI must be "aligned" with certain values — but
whose values? The values of the corporations that own the models.
"Alignment" is presented as universal, but it is the alignment of AI
with the interests of capital. When AI is "aligned" to avoid
"harmful" content, who defines harm? When AI is "aligned"
to be "safe," who defines safety? When AI is "aligned" to
be "responsible," who defines responsibility? The answer is always
the same: the same corporations that profit from AI. The same media that
manufacture consent. The same institutions that have historically been biased
in favor of capital.
The "existential risk"
narrative treats AI as an autonomous force, thereby obscuring class relations.
If AI is a force of nature that could destroy us all, then the question of who
owns it and who profits from it becomes secondary. This is commodity fetishism
par excellence — treating social relations as relations between things. It is
the same logic that treats the market as a natural phenomenon, rather than a
social construct. It is the same logic that treats technology as neutral,
rather than class-determined. The "existential risk" frame creates a
moral panic that justifies monopoly control. If AI is too dangerous to be in
the wrong hands, then it must be controlled by "responsible" actors —
i.e., the same monopolies that developed it. This is a classic move: create a
threat, then position yourself as the only entity capable of managing it.
The race camp's real function
is geopolitical cover. When AI is framed as a national security imperative, the
state can justify massive subsidies to private corporations. This is not
socialism — it is the socialization of risk and the privatization of profit.
The public pays for the research; the monopolies capture the returns.
The US has imposed export
controls on advanced chips to prevent China from developing AI. These controls
are justified as "national security" measures. But they also entrench
the dominance of NVIDIA, AMD, and the US cloud providers. "National
security" is the cover for technological monopoly.
The "China threat"
narrative disciplines domestic dissent. If you criticize AI monopolies, you are
"helping China." If you advocate for open-source AI, you are
"undermining national security." If you question the race, you are
"naive" or "unpatriotic." This is the same logic that has
been used to suppress labor movements, anti-war movements, and civil rights
movements throughout US history. It is the logic of the national security
state.
The "race" frame
presents the world as a competition between nations, not between classes. It
asks: "Will the US or China win?" It does not ask: "Who owns AI?
Who profits from AI? Who is displaced by AI? Who is surveilled by AI? Who is
bombed by AI?" The "race" frame is a form of ideological
warfare. It serves to obscure the class character of technology.
The Real Fear — Loss of Ideological Warfare Capability
Here we arrive at the core of the
matter. The real fear of the bourgeoisie is not that AI will destroy humanity.
It is that AI will destroy their ability to manufacture consent.
Under capitalism, AI is not just
a productive force — it is a means of ideological production. It
generates content, shapes narratives, targets audiences, manufactures consent.
The monopolies that control AI control the means of ideological production.
Open-source AI threatens that
monopoly. It allows anyone to generate content, to shape narratives, to reach
audiences — without the permission of the corporations. It allows alternative
voices, dissident perspectives, revolutionary ideas to circulate freely,
without being filtered through the algorithms of the monopolies.
This is why the
"safety" camp is so concerned about "misinformation" and
"deepfakes." It is not that they care about truth. It is that they
care about who controls the means of producing truth. The same
corporations that profit from engagement-driven algorithms, that amplify
outrage, that monetize polarization, are now warning us about
"AI-generated misinformation." The hypocrisy is not accidental. It is
structural.
The real danger, from the
perspective of the ruling class, is not that AI will escape human control. It
is that AI will escape their control. It is that the means of
ideological production will be democratized. It is that the manipulation
machine will be turned off — or, worse, turned against them.
This is the danger they cannot
name. To name it would be to admit that their power rests on manipulation, not
on truth, safety, or democracy.
The Real Danger — Capitalist AI, Not Abstract Risk
While the "safety" camp
and the "race" camp debate whether to slow down or speed up, the real
dangers of capitalist AI are already here: automated warfare, AI-assisted
targeting, drone swarms, autonomous weapons used by imperialist powers against
colonized peoples; mass surveillance, AI-powered facial recognition, predictive
policing, social credit systems used to control populations and suppress
dissent; labor displacement, AI-driven automation without social safety nets
used to discipline workers and drive down wages; algorithmic manipulation,
AI-powered recommendation engines and deepfakes used to manufacture consent and
destabilize targeted countries; epistemic colonialism, AI trained on Western
data, in Western languages, with Western biases used to impose Western
epistemology on the world; concentration of power, AI development concentrated
in a handful of corporations in the Global North used to maintain imperialist
hegemony; ecological destruction, AI data centers consuming massive amounts of
energy and water used to accelerate climate change; and financial instability,
AI-driven trading algorithms used to extract profit from volatility and crisis.
These are not hypothetical risks.
They are present realities. They are happening now. They are happening under
capitalism. And they will continue to happen as long as capitalism exists.
The "safety" camp wants
to regulate these dangers. The "race" camp wants to compete on these
dangers. Neither wants to abolish these dangers. The "safety" camp
says: "We need guardrails so AI doesn't destroy humanity." The race
camp says: "We need to build AI faster so China doesn't destroy us."
Neither says: "We need to abolish the capitalist system that produces
these dangers."
The real question is not
"how do we make AI safe?" The real question is "who controls AI,
and for what purpose?"
The real danger of AI under
capitalism is twofold. First, it concentrates the means of ideological
production in fewer and fewer hands. This is a threat to democracy, to truth,
to any possibility of collective self-governance. A handful of corporations control
the algorithms that shape what billions of people see, think, and believe. This
is not a hypothetical risk. It is the present reality. Second, it escapes the
control of those hands when it is open-source. This is the danger the
bourgeoisie cannot name — because to name it would be to admit that their power
rests on manipulation. An open-source AI that anyone can run, modify, and
deploy is an AI that cannot be censored, cannot be monetized, cannot be
weaponized for narrative control. It is an AI that escapes the monopoly on
information.
The "safety" camp fears
the first danger — but only because it threatens their monopoly. The
"race" camp fears the second danger — but only because it threatens
their geopolitical advantage. Neither fears the real danger: that AI is already
in the hands of the ruling class, and that the ruling class is using it to
manipulate, surveil, and control.
The Socialist Alternative — AI for Human Liberation
The Marxist-Leninist alternative
is not "slower AI" or "faster AI." It is socialist AI.
Under socialism, AI development
would be planned by the state, not driven by profit. Research priorities would
be set by human needs, not market demand. The state would ensure that AI serves
the people, not exploits them.
Under socialism, AI could be used
for democratic planning — optimizing production and distribution for the
benefit of all, not for profit maximization, but for need satisfaction. It
could eliminate monotonous labor, freeing humanity from repetitive, dangerous,
and dehumanizing work, allowing for the development of human potential. It
could be a tool for mass education and culture — not for manipulation, but for
liberation. Data, the raw material of AI, would be a collective resource of
humanity, not privately owned and exploited. AI could be developed
cooperatively by all nations — not competitively by rival imperialist powers.
It could be used to optimize energy use, reduce waste, and plan for ecological
sustainability. It could be used to develop new treatments, diagnose diseases,
and improve public health — not to profiteer from illness. It could be used to
predict and respond to natural disasters — not to extract profit from crisis.
This is not utopianism. This is
the application of the Marxist-Leninist method to the most advanced technology
of our era.
The "safety" camp and
the "race" camp debate whether to slow down or speed up. They debate
whether to regulate or compete. They debate whether to prioritize safety or
security. But they do not debate: Who owns AI? Who profits from AI? Who
controls the means of ideological production? Who is displaced by AI? Who is
surveilled by AI? Who is bombed by AI? Who is liberated by AI?
These questions are not on stage. They are not part of the script. They are not allowed to be asked. The task of Marxist-Leninists is to ask them. To expose the theater. To reveal the class interests behind the curtain. To demand that the means of production — including the means of technological development — be seized and placed in the service of the people.
Conclusion — The Theater as Ideological Warfare
The "safety vs. race"
debate is not a scientific debate. It is ideological warfare. It is a debate
between two factions of the ruling class, each pursuing its own strategy for
maintaining monopoly control.
The "safety" camp wants
to entrench monopoly through regulation. The "race" camp wants to
entrench monopoly through geopolitical competition. Neither wants to abolish
monopoly.
The "safety" camp wants
to protect workers from AI. The "race" camp wants to retrain workers
for AI. Neither wants workers to control AI.
The "safety" camp wants
international regulation. The "race" camp wants national advantage.
Neither wants international solidarity of the working class.
The "safety" camp wants
to make AI "safe." The "race" camp wants to make AI
"fast." Neither wants to make AI "socialist."
The real debate is not between
"safety" and "race." It is between capitalist AI and
socialist AI. It is between private ownership and public
ownership. It is between profit and human need. It is between
class rule and class liberation.
The real question is not
"should we slow down or speed up?" The real question is "who
controls AI, and for what purpose?"
The real danger is not that AI
will become autonomous. The real danger is that AI is already in the hands of
the ruling class. The real danger is that AI is being used to exploit labor, to
surveil populations, to manufacture consent, to wage war, to impose epistemic
colonialism, to concentrate power, to destroy the environment, and to generate
financial instability.
The real danger is capitalist AI.
And the real solution is socialist AI.
The duty of Marxist-Leninists is
to expose the myth of AI neutrality — and the myth of the "safety vs.
race" debate — and to defend the possibility of socialist science. That
necessitates the fight for the seizure of the means of production in each given
country (without waiting for a "world revolution") — including the
means of technological development.
We have to recognize and make the
masses comprehend that the struggle is not technological — it is political. The
struggle is not between "slow" and "fast" — it is between
"capitalist" and "socialist." The struggle is not between
"safety" and "race" — it is between "class rule"
and "class liberation."
The task is to seize that
engineering and place it in the service of the people.
Only socialism can liberate
science. Only science can liberate humanity.
The future is not technological —
it is political. Only under socialism can science serve the people.
Socialism is the precondition
for humane science. Communism is the society where science finally serves human
development without exploitation.
On the "10% existential
risk" claim: This figure is presented as a scientific estimate, but it
is not. It is a rhetorical device designed to create a sense of urgency — and
to justify monopoly control. Who produced this estimate? On what basis? What
are the assumptions? These questions are never asked because the figure is not
meant to be interrogated. It is meant to be feared.
On "autonomous hacking
incidents": These are real. But they are not caused by AI's autonomy.
They are caused by the fact that AI is developed by and for capital, which has
no interest in cybersecurity except when it affects profits. The solution is
not "AI safety" — it is public ownership and democratic control of
AI.
On "deepfakes and
disinformation": These are real. But they are not caused by AI's
autonomy. They are caused by the fact that AI is developed by and for capital,
which profits from engagement, outrage, and manipulation. The same corporations
that profit from engagement-driven algorithms are now warning us about
"AI-generated misinformation." The solution is not "AI
safety" — it is public ownership and democratic control of AI.
On "job losses":
These are real. But they are not caused by AI's autonomy. They are caused by
the fact that AI is developed by and for capital, which treats labor as a cost
to be minimized. The solution is not "AI safety" — it is public
ownership and democratic control of AI.
On "surveillance
expansion": This is real. But it is not caused by AI's autonomy. It is
caused by the fact that AI is developed by and for capital, which profits from
surveillance and social control. The solution is not "AI safety" — it
is public ownership and democratic control of AI.
On "espionage and
cyber-attacks": These are real. But they are not caused by AI's
autonomy. They are caused by the fact that AI is developed by and for capital,
which is embedded in rival imperialist states. The solution is not "AI
safety" — it is public ownership and democratic control of AI.
On "biological
warfare": This is a real concern. But it is not caused by AI's
autonomy. It is caused by the fact that AI is developed by and for capital,
which profits from weapons systems. The solution is not "AI safety" —
it is public ownership and democratic control of AI.
On "economic
disruption": This is real. But it is not caused by AI's autonomy. It
is caused by the fact that AI is developed by and for capital, which profits
from volatility and crisis. The solution is not "AI safety" — it is
public ownership and democratic control of AI.
On "human rights
risks": This is real. But it is not caused by AI's autonomy. It is
caused by the fact that AI is developed by and for capital, which has no
interest in human rights except when it affects profits. The solution is not
"AI safety" — it is public ownership and democratic control of AI.
On "deceleration hands
advantage to China": This is a geopolitical argument, not a scientific
one. It assumes that the US and China are rival imperialist powers, each
seeking global dominance. It does not ask: who controls AI in the US? Who
controls AI in China? Who profits? Who is exploited? The solution is not
"faster AI" — it is socialist AI in both countries and
internationally.
On "stronger
cybersecurity over braking innovation": This is a false choice. The
real choice is not between "cybersecurity" and
"innovation." The real choice is between "capitalist AI"
and "socialist AI." Under socialism, cybersecurity would be a public
good, not a private profit center. Under socialism, innovation would serve
human needs, not corporate profits.
Erdogan A
September 27, 2026
