Selecionar segmento
Estude com questões de diferentes segmentos
Atenção: Isso limpará todos os campos já preenchidos no filtro!
Foram encontradas 145.674 questões
Resolva questões gratuitamente!
Junte-se a mais de 4 milhões de concurseiros!
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780409
Ética na Administração Pública
Os auditores internos governamentais devem se comportar com
cortesia e respeito no trato com pessoas, mesmo em situações de
divergência de opinião, abstendo-se de emitir juízo ou adotar
práticas que indiquem qualquer tipo de discriminação ou
preconceito.
Assinale a opção que indica o princípio fundamental para a prática da atividade de Auditoria Interna Governamental que exige esse comportamento dos auditores internos.
Assinale a opção que indica o princípio fundamental para a prática da atividade de Auditoria Interna Governamental que exige esse comportamento dos auditores internos.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780408
Auditoria Governamental
Ao realizar um trabalho de auditoria interna, a equipe aplicou as
técnicas de auditoria a seguir.
I. Verificar os registros de documentos ou de ativos, com a finalidade de formar opinião quanto à existência física do objeto examinado;
II. Obter de declaração formal e independente de partes externas a respeito de fatos ligados às operações da Unidade Auditada;
III. Selecionar primeiramente as transações e, em seguida, verificar se existe de fato a documentação que lhe serve de base e, por conseguinte, se aquela transação ocorreu.
Assinale a opção que indica corretamente, na ordem apresentada, as técnicas de auditoria empregadas pela equipe.
I. Verificar os registros de documentos ou de ativos, com a finalidade de formar opinião quanto à existência física do objeto examinado;
II. Obter de declaração formal e independente de partes externas a respeito de fatos ligados às operações da Unidade Auditada;
III. Selecionar primeiramente as transações e, em seguida, verificar se existe de fato a documentação que lhe serve de base e, por conseguinte, se aquela transação ocorreu.
Assinale a opção que indica corretamente, na ordem apresentada, as técnicas de auditoria empregadas pela equipe.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780407
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
The text concludes that the Victorian engineers’ decision
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780406
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
The metaphor used in avoid handing it the keys to the kingdom
(7th paragraph) means avoid giving the system
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780405
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
Introducing in by introducing safety margins (6th paragraph) is
similar in meaning to
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780404
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
The phrase shoddy work costs lives (3rd paragraph) refers to work
that is
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780403
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
The word tricked (1st paragraph) means that LLMs can be
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780402
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
By referring to LLMs as a trendy way of building artificial
intelligence (1st paragraph), the author implies they are
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780401
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
The author compares AI and 19th century engineers to argue that
the latter were
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780400
Inglês
Texto associado
Read the following text and answer the questions.
Artificial Intelligence: The “lethal trifecta”
LARGE LANGUAGE MODELS (LLMs), a trendy way of building
artificial intelligence, have an inherent security problem: they
cannot separate code from data. As a result, they are at risk of a
type of attack called a prompt injection, in which they are tricked
into following commands they should not. Sometimes the result is
merely embarrassing, as when a customer-help agent is persuaded
to talk like a pirate. On other occasions, it is far more damaging.
The worst effects of this flaw are reserved for those who
create what is known as the “lethal trifecta”. If a company, eager
to offer a powerful AI assistant to its employees, gives an LLM
access to untrusted data, the ability to read valuable secrets and
the ability to communicate with the outside world at the same
time, then trouble is sure to follow. And avoiding this is not just a
matter for AI engineers. Ordinary users, too, need to learn how to
use AI safely, because installing the wrong combination of apps
can generate the trifecta accidentally.
Better AI engineering is, though, the first line of defence. And
that means AI engineers need to start thinking like engineers,
who build things like bridges and therefore know that shoddy
work costs lives.
The great works of Victorian England were erected by
engineers who could not be sure of the properties of the materials
they were using. In particular, whether by incompetence or
malfeasance, the iron of the period was often not up to snuff. As a
consequence, engineers erred on the side of caution, overbuilding
to incorporate redundancy into their creations. The result was a
series of centuries-spanning masterpieces.
AI-security providers do not think like this. Conventional
coding is a deterministic practice. Security vulnerabilities are seen
as errors to be fixed, and when fixed, they go away. AI engineers,
inculcated in this way of thinking from their schooldays, therefore
often act as if problems can be solved just with more training
data and more astute system prompts.
These do, indeed, reduce risk. The cleverest frontier models
are better at spotting and refusing malicious requests than their
older or smaller cousins. But they cannot eliminate risk
altogether. Unlike most software, LLMs are probabilistic. Their
output is driven by random selection from likely responses. A
deterministic approach to safety is thus inadequate. A better way
forward is to copy engineers in the physical world and learn to
work with, rather than against, capricious systems that can never
be guaranteed to function as they should. That means becoming
happier dealing with unpredictability by introducing safety
margins, risk tolerance and error rates.
Overbuilding in the AI age might, for instance, mean using a
more powerful model than is needed for the task at hand, to
reduce the risk it will be tricked into doing something
inappropriate. It might mean imposing limits on the number of
queries LLMs can take from external sources, calibrated to the
risk of damage from a malicious query. And mechanical
engineering emphasises failing safely. If an AI system must have
access to secrets, then avoid handing it the keys to the kingdom.
In the physical world, bridges have weight limits – even if
they are not always stated clearly to drivers. And, importantly,
these are well within the actual tolerances that calculations
suggest a bridge will bear. The time has now come for the virtual
world of AI systems to be similarly equipped.
Adapted from The Economist, September 27th, 2025, p. 10
Based on the text, mark the statements below as true (T) or false (F).
I. AI models are watertight when it comes to safety risks.
II Bridges built in the Victorian Age were proven to be quite fragile.
III. A deterministic model does not deal with randomness.
The statements are, respectively,
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780399
Português
Assinale a opção que apresenta o texto que deve ser incluído
entre os narrativos.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780398
Português
Assinale a opção em que se cometeu um erro no emprego de
palavras parônimas ou homônimas.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780397
Português
As opções a seguir mostram orações adjetivas sublinhadas.
Assinale a opção que apresenta a substituição adequada de uma
oração por um adjetivo.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780396
Português
Nas opções a seguir há termos destacados que são referidos de
forma diferente na continuidade do texto.
Assinale a opção em que essa referência é feita por um termo geral.
Assinale a opção em que essa referência é feita por um termo geral.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780395
Português
As opções a seguir apresentam cinco exemplos de inícios de narrativas literárias.
Assinale a que mostra um início em que há preocupação com a identificação das realidades referidas.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780394
Português
Leia o texto a seguir.
Belo Horizonte, a capital de Minas Gerais, foi palco de um brutal crime transfóbico, registrado pelas câmeras de segurança na região de Venda Nova.
Nas imagens, é possível ver Christina Maciel Oliveira, de 45 anos, caminhando tranquilamente quando, de repente, é golpeada por Matheus Henrique Santos Rodrigues, de 24 anos.
A mulher é derrubada por Matheus, que, em seguida, inicia uma sequência de chutes e pisadas na cabeça de Christina, que desmaia. Após o espancamento, o homem pega seu chinelo e sai caminhando tranquilamente.
Pessoas que presenciaram o crime acionaram o SAMU e a polícia. Os agentes de socorro tentaram reanimar Christina, mas ela não resistiu e morreu no local.
Matheus Henrique Santos Rodrigues foi localizado a poucos metros da cena do crime e preso em flagrante pelos policiais.
Fórum 21/10/2025.
Sobre o processo de referenciação no texto, assinale a afirmativa correta.
Belo Horizonte, a capital de Minas Gerais, foi palco de um brutal crime transfóbico, registrado pelas câmeras de segurança na região de Venda Nova.
Nas imagens, é possível ver Christina Maciel Oliveira, de 45 anos, caminhando tranquilamente quando, de repente, é golpeada por Matheus Henrique Santos Rodrigues, de 24 anos.
A mulher é derrubada por Matheus, que, em seguida, inicia uma sequência de chutes e pisadas na cabeça de Christina, que desmaia. Após o espancamento, o homem pega seu chinelo e sai caminhando tranquilamente.
Pessoas que presenciaram o crime acionaram o SAMU e a polícia. Os agentes de socorro tentaram reanimar Christina, mas ela não resistiu e morreu no local.
Matheus Henrique Santos Rodrigues foi localizado a poucos metros da cena do crime e preso em flagrante pelos policiais.
Fórum 21/10/2025.
Sobre o processo de referenciação no texto, assinale a afirmativa correta.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780393
Português
Assinale a frase que mostra uma construção correta.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780392
Português
Leia o seguinte texto, retirado de uma história infantil:
Era uma vez um poderoso rei chamado Uriel, que vivia num majestoso castelo ao norte do reino de Mantrifás. O rei tinha uma corte de valorosos cavaleiros que se reuniam numa sala de cristal toda vez que havia de deliberar-se sobre os assuntos mais importantes de um reino.
Sobre a significação ou a estruturação desse fragmento textual, assinale a afirmativa incorreta.
Era uma vez um poderoso rei chamado Uriel, que vivia num majestoso castelo ao norte do reino de Mantrifás. O rei tinha uma corte de valorosos cavaleiros que se reuniam numa sala de cristal toda vez que havia de deliberar-se sobre os assuntos mais importantes de um reino.
Sobre a significação ou a estruturação desse fragmento textual, assinale a afirmativa incorreta.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780391
Português
Leia o texto a seguir.
Diante da agência dos Correios havia uma oficina de carpintaria e uma loja de roupa infantil. Tanto a oficina como a loja estavam afetadas pelo plano de urbanização da zona, mas aquela, não estava previsto derrubá-la, até meados de abril.
Em relação à significação ou à estruturação desse fragmento textual, assinale a afirmativa inadequada.
Diante da agência dos Correios havia uma oficina de carpintaria e uma loja de roupa infantil. Tanto a oficina como a loja estavam afetadas pelo plano de urbanização da zona, mas aquela, não estava previsto derrubá-la, até meados de abril.
Em relação à significação ou à estruturação desse fragmento textual, assinale a afirmativa inadequada.
Ano: 2025
Banca:
FGV
Órgão:
CGE-SP
Provas:
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Auditoria
|
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Tecnologia da Informação |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Correição e Combate à Corrupção |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Obras e Concessões |
FGV - 2025 - CGE-SP - Auditor Estadual de Controle - Contabilidade Pública e Finanças |
Q3780390
Português
Leia o texto a seguir.
Os pés do homem se afundaram na areia, deixando uma marca informe, como se fosse a pegada de algum animal. Treparam sobre as pedras, usando as unhas ao sentirem a inclinação da subida, logo caminharam para cima, buscando o horizonte.
‘Pés chatos’ – disse o que o seguia. E um dedo de menos. Falta-lhe o dedo central no pé esquerdo. Não abundam indivíduos com essa marca. Assim será fácil.
A estranheza presente nesse texto, que é o início de um romance, decorre
Os pés do homem se afundaram na areia, deixando uma marca informe, como se fosse a pegada de algum animal. Treparam sobre as pedras, usando as unhas ao sentirem a inclinação da subida, logo caminharam para cima, buscando o horizonte.
‘Pés chatos’ – disse o que o seguia. E um dedo de menos. Falta-lhe o dedo central no pé esquerdo. Não abundam indivíduos com essa marca. Assim será fácil.
A estranheza presente nesse texto, que é o início de um romance, decorre
Conheça nossos planos